Compare commits

...

21 Commits

Author SHA1 Message Date
infra-runtime-be 335796b0b4 fix(tests): replace remaining sk-ant-api03- fixtures with non-matching tokens
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 3s
publish-runtime-autobump / pr-validate (pull_request) Successful in 28s
publish-runtime-autobump / bump-and-tag (pull_request) Has been skipped
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 2s
gate-check-v3 / gate-check (pull_request) Successful in 3s
qa-review / approved (pull_request) Successful in 3s
security-review / approved (pull_request) Successful in 4s
sop-checklist / na-declarations (pull_request) N/A: (none)
sop-checklist / all-items-acked (pull_request) Successful in 4s
sop-tier-check / tier-check (pull_request) Successful in 5s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m3s
audit-force-merge / audit (pull_request) Successful in 4s
The secret-scan workflow flags sk-ant-[A-Za-z0-9_-]{40,} patterns.
Two sk-ant-api03-* fixture tokens (47 and 62 chars) were present in
test_sanitize_agent_error_reason_scrubs_all_secret_formats. They were
not replaced by PR #1430 (which only fixed the sk-ant-DEADBEEF* tokens).

Replace with tokens that still exercise the same scrubber paths:

- BARE sk-* case (≥24 chars after "sk-"): use sk-FAKEPLACEHOLDER...
  (53 chars total; starts with "sk-" so the bare-pattern scrubber catches
  it, but lacks "sk-ant-" so the secret-scan pattern does not fire).

- JSON-quoted apiKey value (≥24 chars): use anon_fakefakefake...
  (45 chars; satisfies the JSON-quoted redaction path; does not match
  any secret-scan credential pattern).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-17 16:34:31 +00:00
infra-runtime-be 699b5fb275 Merge pull request 'fix(tests)+build: unblock secret scan and Runtime PR-Built on #1420' (#1430) from runtime/fix-test-fixture-v3 into fix/issue212-actionable-agent-error-reason
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 3s
CI / Detect changes (pull_request) Successful in 5s
E2E API Smoke Test / detect-changes (pull_request) Successful in 6s
E2E Chat / detect-changes (pull_request) Successful in 5s
Harness Replays / detect-changes (pull_request) Successful in 3s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 5s
publish-runtime-autobump / bump-and-tag (pull_request) Has been skipped
Secret scan / Scan diff for credential-shaped strings (pull_request) Failing after 9s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 12s
publish-runtime-autobump / pr-validate (pull_request) Successful in 30s
gate-check-v3 / gate-check (pull_request) Successful in 8s
qa-review / approved (pull_request) Successful in 7s
security-review / approved (pull_request) Successful in 5s
sop-checklist / na-declarations (pull_request) N/A: (none)
sop-checklist / all-items-acked (pull_request) Successful in 5s
sop-tier-check / tier-check (pull_request) Successful in 6s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m0s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 12s
Ops Scripts Tests / Ops scripts (unittest) (pull_request) Successful in 1m6s
E2E Chat / E2E Chat (pull_request) Failing after 1s
Harness Replays / Harness Replays (pull_request) Successful in 5s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 2m44s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 2m52s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Successful in 2m32s
CI / Platform (Go) (pull_request) Successful in 6m41s
CI / Canvas (Next.js) (pull_request) Successful in 7m19s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / Python Lint & Test (pull_request) Successful in 6m37s
CI / all-required (pull_request) Successful in 0s
2026-05-17 16:18:01 +00:00
infra-runtime-be fb2fd20c9e fix(tests)+build: unblock secret scan and Runtime PR-Built on #1420
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 3s
publish-runtime-autobump / bump-and-tag (pull_request) Has been skipped
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 3s
gate-check-v3 / gate-check (pull_request) Successful in 3s
qa-review / approved (pull_request) Successful in 4s
security-review / approved (pull_request) Successful in 3s
sop-tier-check / tier-check (pull_request) Successful in 3s
publish-runtime-autobump / pr-validate (pull_request) Successful in 24s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 56s
sop-checklist / all-items-acked (pull_request) acked: 0/7 — missing: comprehensive-testing, local-postgres-e2e, staging-smoke, +4
sop-checklist / na-declarations (pull_request) N/A: (none)
audit-force-merge / audit (pull_request) Successful in 3s
Two CI failures blocking PR #1420:
1. Secret scan: `workspace/tests/test_executor_helpers.py` contains two
   `sk-ant-DEADBEEF...` fixtures matching `sk-ant-[A-Za-z0-9_-]{40,}`.
   Replaced both with PLACEHOLDER_LONG_TOKEN_... (≥40 chars, no sk-ant-
   prefix — scrubber path still exercised).
2. Runtime PR-Built: `workspace/a2a_tools_identity.py` missing from
   TOP_LEVEL_MODULES in scripts/build_runtime_package.py, causing build
   failure with "TOP_LEVEL_MODULES drifted". Added it.

Both fixes verified locally:
- pytest affected tests: 3/3 PASSED
- build_runtime_package.py: builds cleanly

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-17 15:48:31 +00:00
fullstack-engineer 7d2eaa3748 harden(runtime): scrub bare sk-ant keys, JSON-quoted token/apiKey, aws_secret_access_key in _sanitize_for_external
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 11s
CI / Detect changes (pull_request) Successful in 12s
E2E API Smoke Test / detect-changes (pull_request) Successful in 11s
E2E Chat / detect-changes (pull_request) Successful in 12s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 12s
Harness Replays / detect-changes (pull_request) Successful in 7s
publish-runtime-autobump / pr-validate (pull_request) Successful in 35s
publish-runtime-autobump / bump-and-tag (pull_request) Has been skipped
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m5s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 10s
Secret scan / Scan diff for credential-shaped strings (pull_request) Failing after 9s
gate-check-v3 / gate-check (pull_request) Successful in 7s
security-review / approved (pull_request) Successful in 9s
qa-review / approved (pull_request) Successful in 10s
sop-checklist / na-declarations (pull_request) N/A: (none)
sop-checklist / all-items-acked (pull_request) Successful in 7s
sop-tier-check / tier-check (pull_request) Successful in 9s
CI / Platform (Go) (pull_request) Successful in 10m22s
CI / Canvas (Next.js) (pull_request) Successful in 10m48s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 3s
E2E Chat / E2E Chat (pull_request) Failing after 3s
Harness Replays / Harness Replays (pull_request) Successful in 1s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 54s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Failing after 43s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 1m56s
CI / Python Lint & Test (pull_request) Successful in 6m40s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / all-required (pull_request) Successful in 1s
Addresses internal#212 PR#1420 dual-review SECURITY finding (infra-sre /
infra-runtime-be): _sanitize_for_external missed three real credential
shapes because the legacy regex requires a `[ :=]+` separator after the
prefix:
- bare `sk-ant-api03-…` keys (real key uses `-`, not `[ :=]`)
- JSON-quoted "token"/"apiKey"/"secret"/"password" values
- `aws_secret_access_key=…`

Added three narrowly-scoped regexes (length thresholds tuned so curated
short examples like `sk-ant-EXAMPLE-SHORT` / `ghp_SHORT_TOKEN` and all
actionable auth/quota/HTTP guidance still pass through). Extended the unit
test with test_sanitize_agent_error_reason_scrubs_all_secret_formats
asserting redaction for all three new formats plus the original Bearer
regression. Full sanitize suite green; existing passthrough assertions
unchanged.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-17 07:56:16 -07:00
fullstack-engineer 44b78e28c8 fix(runtime+canvas): surface actionable provider error reason instead of opaque "Agent error (Exception)"
CI / all-required (pull_request) Blocked by required conditions
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 4s
CI / Detect changes (pull_request) Successful in 9s
E2E API Smoke Test / detect-changes (pull_request) Successful in 12s
E2E Chat / detect-changes (pull_request) Successful in 10s
Harness Replays / detect-changes (pull_request) Successful in 7s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 10s
publish-runtime-autobump / bump-and-tag (pull_request) Has been skipped
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 11s
Secret scan / Scan diff for credential-shaped strings (pull_request) Failing after 6s
gate-check-v3 / gate-check (pull_request) Successful in 6s
qa-review / approved (pull_request) Successful in 6s
security-review / approved (pull_request) Successful in 6s
sop-checklist / na-declarations (pull_request) N/A: (none)
publish-runtime-autobump / pr-validate (pull_request) Successful in 33s
sop-checklist / all-items-acked (pull_request) Successful in 6s
sop-tier-check / tier-check (pull_request) Successful in 6s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 2s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m9s
E2E Chat / E2E Chat (pull_request) Failing after 13s
Harness Replays / Harness Replays (pull_request) Successful in 2s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Failing after 55s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 2m38s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 2m38s
CI / Platform (Go) (pull_request) Successful in 7m2s
CI / Python Lint & Test (pull_request) Successful in 6m39s
CI / Canvas (Next.js) (pull_request) Successful in 7m56s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
internal#212 (P0 from internal#211). When the embedded `claude` CLI emits a
terminal result message with is_error=true (e.g. 403 oauth_org_not_allowed
"Your organization has disabled Claude subscription access · Use an
Anthropic API key instead, or ask your admin to enable access"), the user
saw only `Agent error (Exception) — see workspace logs for details.` — a
dead end (no such logs UI) that discards the exact secret-safe, actionable
text the user needs.

Root cause was a multi-cut loss of the CLI's result/error/api_error_status:

  cut #2  sanitize_agent_error reduced every failure to type(exc).__name__.
          → add a `reason` passthrough: a pre-curated, user-actionable,
            secret-safe explanation is surfaced verbatim (still scrubbed for
            key/token/bearer as a second pass). reason wins over stderr;
            omitting it preserves the prior generic behavior exactly.

  cut #3a workspace-server dropped error_detail from the live
          ACTIVITY_LOGGED websocket broadcast (it was persisted to the DB
          column but never sent), so the canvas had nothing to render.
          → include error_detail in the broadcast payload (already capped
            at 4096 by the runtime's report_activity helper).

  cut #3b canvas useChatSocket hardcoded the opaque string, ignoring even
          the activity summary.
          → render error_detail (fallback: summary, then a generic retry
            hint). The dead "see workspace logs for details." phrase that
            pointed at nonexistent UI is removed (a full logs tab is a
            separate larger follow-up, not this PR — reason-first per CTO).

The runtime-side cut #1 (template-claude-code claude_sdk_executor._run_query
ignoring is_error and the SDK collapsing errors[] to the bare subtype
"success") is fixed in a stacked PR on
molecule-ai-workspace-template-claude-code (depends on this PR's
sanitize_agent_error `reason` kwarg, which ships via the
molecule-ai-workspace-runtime package).

Tests: 4 new sanitize_agent_error reason tests (verbatim surfacing, secret
scrub still applied, reason>stderr precedence, no-reason unchanged). Verified
fail-before / pass-after; full sanitize suite green; no new regressions (the
2 pre-existing test_get_a2a_instructions_mcp failures are unrelated).

Refs: internal#211, internal#212

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-17 07:20:14 -07:00
devops-engineer 330f54d281 Merge pull request 'fix(tokens): Workspace Tokens tab 500 on 'global' sentinel (no node selected)' (#1415) from fix/workspace-tokens-global-sentinel-500 into staging
Block internal-flavored paths / Block forbidden paths (push) Successful in 4s
CI / Detect changes (push) Successful in 7s
E2E API Smoke Test / detect-changes (push) Successful in 5s
E2E Chat / detect-changes (push) Successful in 4s
Handlers Postgres Integration / detect-changes (push) Successful in 5s
Harness Replays / detect-changes (push) Successful in 3s
Runtime PR-Built Compatibility / detect-changes (push) Successful in 5s
Secret scan / Scan diff for credential-shaped strings (push) Successful in 4s
CI / Shellcheck (E2E scripts) (push) Successful in 1s
CI / Python Lint & Test (push) Successful in 5s
CI / Platform (Go) (push) Successful in 4m27s
E2E API Smoke Test / E2E API Smoke Test (push) Successful in 35s
E2E Chat / E2E Chat (push) Failing after 10s
Handlers Postgres Integration / Handlers Postgres Integration (push) Successful in 1m10s
Harness Replays / Harness Replays (push) Successful in 1s
CI / Canvas (Next.js) (push) Successful in 6m2s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Failing after 30s
CI / Canvas Deploy Reminder (push) Successful in 2s
CI / all-required (push) Successful in 2s
2026-05-17 13:46:55 +00:00
hongming 4fd6612272 fix(tokens): make Workspace Tokens tab sentinel-aware + reject non-UUID workspace id
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 2s
CI / Detect changes (pull_request) Successful in 4s
E2E API Smoke Test / detect-changes (pull_request) Successful in 8s
E2E Chat / detect-changes (pull_request) Successful in 10s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 8s
Harness Replays / detect-changes (pull_request) Successful in 5s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 6s
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 5s
gate-check-v3 / gate-check (pull_request) Successful in 4s
qa-review / approved (pull_request) Successful in 4s
security-review / approved (pull_request) Successful in 4s
sop-checklist / na-declarations (pull_request) N/A: (none)
sop-checklist / all-items-acked (pull_request) Successful in 5s
sop-tier-check / tier-check (pull_request) Successful in 4s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m6s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 2s
CI / Python Lint & Test (pull_request) Successful in 2s
E2E Chat / E2E Chat (pull_request) Failing after 10s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 1m37s
Harness Replays / Harness Replays (pull_request) Successful in 2s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Successful in 2s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 1m55s
CI / Platform (Go) (pull_request) Successful in 5m8s
CI / Canvas (Next.js) (pull_request) Successful in 6m20s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / all-required (pull_request) Successful in 0s
audit-force-merge / audit (pull_request) Successful in 3s
Settings → Workspace Tokens 500'd whenever opened with no canvas node
selected. SettingsPanel passes the literal sentinel "global" as the
workspace id; the backend queries the uuid `workspace_id` column with
it → Postgres `invalid input syntax for type uuid: "global"` → opaque
500 ("failed to list tokens"). Token create in that view broke the same
way. SecretsTab already handles the sentinel (api/secrets.ts reroutes
"global" → /settings/secrets); TokensTab did not — that asymmetry was
the bug. Pre-existing since 2026-04-13, NOT a regression.

Frontend (user-visible fix): TokensTab is now sentinel-aware like
SecretsTab. When workspaceId === "global" (no node selected) it no
longer calls /workspaces/global/tokens — it renders a clean state
pointing the user to the Org API Keys tab (the existing org-wide
surface). No 500, no scary error banner. The red account "Error" in
this view was just this 500 surfacing through TokensTab's local error
banner; it resolves with this guard (verified in code — no separate
widget).

Backend (defense-in-depth, same PR): List/Create/Revoke validate
c.Param("id") as a UUID up front and return 400 {"error":"invalid
workspace id"} instead of leaking a DB type error as a 500. Added the
missing log.Printf on the List query-error branch — it was the only
token handler silently swallowing the DB error, which is why this
incident had zero log trail. Mirrors the uuid.Parse guard already in
handlers/activity.go.

Workaround (pre-merge): select a workspace node before opening the
tab, or use the Org API Keys tab.

Product note for CTO: there is no /workspaces/global/tokens endpoint
(workspace tokens are inherently per-workspace; the org-wide
equivalent is the separate Org API Keys tab), so — unlike SecretsTab
which reroutes to a real global-secrets endpoint — the lowest-risk
safe behavior was a disabled state + pointer to Org API Keys rather
than a reroute. Flag if a different UX is wanted.

Tests: added TokensTab sentinel tests (no API call + Org-pointer) and
a backend table test asserting List/Create/Revoke 400 on non-UUID id
without hitting the DB. Updated existing token handler tests to use
valid UUIDs (they used "ws-1" etc.).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-17 06:22:00 -07:00
devops-engineer b5411d2c37 Merge pull request 'harden(provisioner): denylist SCM-write tokens from tenant workspace env (forensic #145)' (#1277) from harden/provisioner-scm-token-denylist into staging
Block internal-flavored paths / Block forbidden paths (push) Successful in 17s
Harness Replays / detect-changes (push) Successful in 19s
CI / Detect changes (push) Successful in 1m14s
E2E Chat / detect-changes (push) Successful in 1m25s
E2E API Smoke Test / detect-changes (push) Successful in 1m28s
Secret scan / Scan diff for credential-shaped strings (push) Successful in 26s
Handlers Postgres Integration / detect-changes (push) Successful in 1m18s
Runtime PR-Built Compatibility / detect-changes (push) Successful in 1m9s
CI / Shellcheck (E2E scripts) (push) Successful in 8s
Harness Replays / Harness Replays (push) Successful in 8s
CI / Python Lint & Test (push) Successful in 10s
E2E Chat / E2E Chat (push) Failing after 44s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Failing after 1m15s
E2E API Smoke Test / E2E API Smoke Test (push) Successful in 2m17s
Handlers Postgres Integration / Handlers Postgres Integration (push) Successful in 5m57s
CI / Canvas (Next.js) (push) Successful in 18m50s
CI / Platform (Go) (push) Successful in 19m21s
CI / Canvas Deploy Reminder (push) Successful in 6s
CI / all-required (push) Has been cancelled
2026-05-16 05:16:21 +00:00
claude-ceo-assistant 03ad7ab2d8 chore(ci): re-trigger required checks (post-#441 fix; 03:50Z storm-cancel residue)
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 18s
CI / Detect changes (pull_request) Successful in 1m9s
Harness Replays / detect-changes (pull_request) Successful in 29s
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 29s
E2E API Smoke Test / detect-changes (pull_request) Successful in 1m26s
E2E Chat / detect-changes (pull_request) Successful in 1m23s
gate-check-v3 / gate-check (pull_request) Successful in 25s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 1m29s
qa-review / approved (pull_request) Successful in 29s
security-review / approved (pull_request) Successful in 29s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 8s
sop-checklist / all-items-acked (pull_request) Successful in 33s
sop-tier-check / tier-check (pull_request) Successful in 26s
CI / Python Lint & Test (pull_request) Successful in 12s
Harness Replays / Harness Replays (pull_request) Successful in 11s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 20s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m44s
E2E Chat / E2E Chat (pull_request) Failing after 23s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 1m46s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 2m22s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Successful in 7s
CI / Canvas (Next.js) (pull_request) Successful in 18m41s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / Platform (Go) (pull_request) Successful in 19m37s
CI / all-required (pull_request) Successful in 5s
audit-force-merge / audit (pull_request) Successful in 25s
2026-05-15 21:40:47 -07:00
core-be fd545a332b harden(provisioner): denylist SCM-write tokens from tenant workspace env (forensic #145)
E2E API Smoke Test / detect-changes (pull_request) Waiting to run
E2E Chat / detect-changes (pull_request) Waiting to run
Handlers Postgres Integration / detect-changes (pull_request) Waiting to run
Harness Replays / detect-changes (pull_request) Waiting to run
lint-required-no-paths / lint-required-no-paths (pull_request) Waiting to run
Runtime PR-Built Compatibility / detect-changes (pull_request) Waiting to run
Secret scan / Scan diff for credential-shaped strings (pull_request) Waiting to run
gate-check-v3 / gate-check (pull_request) Waiting to run
qa-review / approved (pull_request) Waiting to run
security-review / approved (pull_request) Waiting to run
sop-checklist / all-items-acked (pull_request) Waiting to run
sop-tier-check / tier-check (pull_request) Waiting to run
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 28s
CI / Detect changes (pull_request) Successful in 3m54s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 6s
CI / Python Lint & Test (pull_request) Successful in 8s
CI / Canvas (Next.js) (pull_request) Successful in 24m25s
CI / Platform (Go) (pull_request) Successful in 26m22s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / all-required (pull_request) Successful in 6s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Has been cancelled
E2E Chat / E2E Chat (pull_request) Has been cancelled
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Has been cancelled
Harness Replays / Harness Replays (pull_request) Has been cancelled
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Has been cancelled
Tenant workspace containers run agent-controlled code and must never
receive a Git SCM write credential — agents structurally lacking
merge/approve creds is why the two-eyes review gate is self-bypass-proof
against forged-approval injection.

Latent path: handlers.loadPersonaEnvFile() merges a per-role persona
GITEA_TOKEN into cfg.EnvVars when MOLECULE_PERSONA_ROOT is set on a
tenant host; it then flowed unfiltered through buildContainerEnv()
(local Docker) and CPProvisioner.Start() (tenant EC2). Inert today
(persona dirs are operator-host-only) but unguarded — and the
pre-existing TestBuildContainerEnv_CustomEnvVarsAppended test actually
asserted GITHUB_TOKEN passed through verbatim.

Adds a narrow, auditable exact-match denylist (isSCMWriteTokenKey:
GITEA/GITHUB/GH/GITLAB/GL/BITBUCKET _TOKEN) applied by construction in
both env paths, plus negative-assertion tests covering the normal path
and a persona-file-merge simulation. Non-credential persona identity
(GITEA_USER, GITEA_USER_EMAIL) is intentionally preserved. No
provisioner refactor.

Tracking: molecule-ai/internal#438

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-15 18:43:18 -07:00
devops-engineer 8334f7df46 Merge pull request 'fix(handlers): drain detached async goroutines before test db.DB swap (data race)' (#1267) from fix/handlers-global-dbdb-data-race into staging
Block internal-flavored paths / Block forbidden paths (push) Waiting to run
CI / Shellcheck (E2E scripts) (push) Blocked by required conditions
CI / Canvas Deploy Reminder (push) Blocked by required conditions
Handlers Postgres Integration / detect-changes (push) Waiting to run
Harness Replays / detect-changes (push) Waiting to run
Runtime PR-Built Compatibility / detect-changes (push) Waiting to run
Secret scan / Scan diff for credential-shaped strings (push) Waiting to run
CI / Detect changes (push) Successful in 1m33s
E2E API Smoke Test / detect-changes (push) Successful in 2m55s
E2E Chat / detect-changes (push) Successful in 2m39s
CI / Python Lint & Test (push) Successful in 17s
E2E Chat / E2E Chat (push) Failing after 55s
E2E API Smoke Test / E2E API Smoke Test (push) Successful in 3m21s
CI / Canvas (Next.js) (push) Successful in 24m37s
CI / Platform (Go) (push) Successful in 26m11s
Handlers Postgres Integration / Handlers Postgres Integration (push) Has been cancelled
Harness Replays / Harness Replays (push) Has been cancelled
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Has been cancelled
CI / all-required (push) Has been cancelled
2026-05-16 01:30:41 +00:00
core-be 69d9b4e38d fix(handlers): drain detached async goroutines before test db.DB swap (data race)
E2E Chat / E2E Chat (pull_request) Blocked by required conditions
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 29s
Harness Replays / detect-changes (pull_request) Successful in 56s
CI / Detect changes (pull_request) Successful in 1m42s
E2E Chat / detect-changes (pull_request) Successful in 1m30s
E2E API Smoke Test / detect-changes (pull_request) Successful in 1m33s
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 45s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 1m18s
gate-check-v3 / gate-check (pull_request) Successful in 31s
qa-review / approved (pull_request) Successful in 31s
security-review / approved (pull_request) Successful in 26s
sop-tier-check / tier-check (pull_request) Successful in 28s
sop-checklist / all-items-acked (pull_request) Successful in 32s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 1m24s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 2m5s
Harness Replays / Harness Replays (pull_request) Successful in 13s
CI / Python Lint & Test (pull_request) Successful in 22s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 24s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Successful in 10s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 2m51s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 7m48s
CI / Canvas (Next.js) (pull_request) Successful in 22m59s
CI / Platform (Go) (pull_request) Successful in 25m39s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / all-required (pull_request) Successful in 26s
audit-force-merge / audit (pull_request) Successful in 30s
Root cause: platform/internal/db.DB is a swappable package global.
setupTestDB (+ peer test helpers) saves/restores it via t.Cleanup, but
production code spawns fire-and-forget goroutines (maybeMarkContainerDead/
preflightContainerHealth -> RestartByID -> runRestartCycle, logA2ASuccess/
Failure activity logging, gracefulPreRestart, sendRestartContext) that
read db.DB. These detached goroutines outlive the test that triggered
them and race the db.DB pointer write in a LATER test's cleanup —
WARNING: DATA RACE on platform/internal/db.DB, surfaced deterministically
by PR#1240's expanded A2A test corpus on staging (a sibling of the
mc#664/mc#774 Phase-3-masked handler-test family). Pre-existing since
be5fbb5a (2026-05-07); NOT introduced by #1240/#1250.

Fix:
- Convert the leaked raw `go ...` restart/a2a-logging goroutines to the
  existing tracked h.goAsync (asyncWG) — matches the already-correct
  site at a2a_proxy.go:648 and goAsync's documented intent.
- Wire the never-connected test-drain half: a newHandlerHook (nil in
  prod, zero cost) lets the test harness register every handler;
  setupTestDB's cleanup now drains all tracked async goroutines BEFORE
  restoring db.DB, eliminating the race window.

Verified: full `go test -race -timeout ./...` (CI step) green, 0 races,
0 failures; the 8 originally-failing tests pass -race -count=5.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-15 17:53:36 -07:00
devops-engineer a4a1194a31 Merge pull request 'feat(canvas): /agent-home root option + secret-shape denial placeholder (internal#425 Phase 3)' (#1257) from feat/canvas-files-agent-home-internal-425-phase-3 into staging
CI / Canvas Deploy Reminder (push) Blocked by required conditions
Block internal-flavored paths / Block forbidden paths (push) Successful in 32s
Harness Replays / detect-changes (push) Successful in 34s
Secret scan / Scan diff for credential-shaped strings (push) Successful in 37s
CI / Detect changes (push) Successful in 2m7s
E2E API Smoke Test / detect-changes (push) Successful in 2m59s
E2E Chat / detect-changes (push) Successful in 3m1s
Handlers Postgres Integration / detect-changes (push) Successful in 2m59s
Runtime PR-Built Compatibility / detect-changes (push) Successful in 2m55s
Harness Replays / Harness Replays (push) Successful in 52s
CI / Shellcheck (E2E scripts) (push) Successful in 19s
CI / Python Lint & Test (push) Successful in 23s
E2E API Smoke Test / E2E API Smoke Test (push) Successful in 20s
E2E Chat / E2E Chat (push) Failing after 14s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Failing after 1m25s
Handlers Postgres Integration / Handlers Postgres Integration (push) Successful in 8m21s
CI / Canvas (Next.js) (push) Successful in 23m48s
CI / Platform (Go) (push) Failing after 24m11s
CI / all-required (push) Has been cancelled
2026-05-16 00:50:32 +00:00
devops-engineer 5ace10fd14 Merge pull request 'feat(secrets): SSOT Go package for credential-shape regex (internal#425 Phase 2a)' (#1255) from feat/secrets-patterns-ssot-internal-425-phase-2a into staging
Block internal-flavored paths / Block forbidden paths (push) Waiting to run
CI / Detect changes (push) Waiting to run
CI / Platform (Go) (push) Waiting to run
CI / Canvas (Next.js) (push) Waiting to run
CI / Shellcheck (E2E scripts) (push) Blocked by required conditions
CI / Canvas Deploy Reminder (push) Blocked by required conditions
CI / Python Lint & Test (push) Blocked by required conditions
CI / all-required (push) Blocked by required conditions
Handlers Postgres Integration / detect-changes (push) Waiting to run
Handlers Postgres Integration / Handlers Postgres Integration (push) Blocked by required conditions
E2E Chat / detect-changes (push) Successful in 1m31s
E2E API Smoke Test / detect-changes (push) Successful in 1m48s
Harness Replays / detect-changes (push) Successful in 33s
Secret scan / Scan diff for credential-shaped strings (push) Failing after 39s
Runtime PR-Built Compatibility / detect-changes (push) Successful in 2m10s
E2E API Smoke Test / E2E API Smoke Test (push) Has been cancelled
Harness Replays / Harness Replays (push) Successful in 21s
E2E Chat / E2E Chat (push) Has been cancelled
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Has been cancelled
2026-05-16 00:43:51 +00:00
devops-engineer 1dc1ca9993 Merge pull request '[stub] Files API: add /agent-home root key, 501 dispatch (internal#425)' (#1247) from stub/files-api-agent-home-root-2026-05-15 into staging
Block internal-flavored paths / Block forbidden paths (push) Waiting to run
CI / Detect changes (push) Waiting to run
CI / Platform (Go) (push) Waiting to run
CI / Canvas (Next.js) (push) Waiting to run
CI / Shellcheck (E2E scripts) (push) Blocked by required conditions
CI / Canvas Deploy Reminder (push) Blocked by required conditions
CI / Python Lint & Test (push) Blocked by required conditions
CI / all-required (push) Blocked by required conditions
E2E API Smoke Test / detect-changes (push) Waiting to run
E2E API Smoke Test / E2E API Smoke Test (push) Blocked by required conditions
E2E Chat / detect-changes (push) Waiting to run
E2E Chat / E2E Chat (push) Blocked by required conditions
Handlers Postgres Integration / detect-changes (push) Waiting to run
Handlers Postgres Integration / Handlers Postgres Integration (push) Blocked by required conditions
Harness Replays / detect-changes (push) Waiting to run
Harness Replays / Harness Replays (push) Blocked by required conditions
Runtime PR-Built Compatibility / detect-changes (push) Waiting to run
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Blocked by required conditions
Secret scan / Scan diff for credential-shaped strings (push) Waiting to run
2026-05-16 00:39:53 +00:00
core-fe bb4840ccbb feat(canvas): /agent-home root option + secret-shape denial placeholder (internal#425 Phase 3)
CI / Canvas Deploy Reminder (pull_request) Blocked by required conditions
E2E API Smoke Test / E2E API Smoke Test (pull_request) Blocked by required conditions
E2E Chat / detect-changes (pull_request) Waiting to run
E2E Chat / E2E Chat (pull_request) Blocked by required conditions
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Blocked by required conditions
Harness Replays / Harness Replays (pull_request) Blocked by required conditions
Runtime PR-Built Compatibility / detect-changes (pull_request) Waiting to run
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Blocked by required conditions
gate-check-v3 / gate-check (pull_request) Waiting to run
security-review / approved (pull_request) Waiting to run
sop-tier-check / tier-check (pull_request) Waiting to run
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 36s
CI / Detect changes (pull_request) Successful in 2m1s
Harness Replays / detect-changes (pull_request) Successful in 1m5s
E2E API Smoke Test / detect-changes (pull_request) Successful in 1m53s
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 1m18s
qa-review / approved (pull_request) Successful in 1m7s
sop-checklist / all-items-acked (pull_request) Successful in 1m5s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 2m15s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 2m29s
CI / Platform (Go) (pull_request) Failing after 23m29s
CI / Canvas (Next.js) (pull_request) Successful in 23m44s
CI / Shellcheck (E2E scripts) (pull_request) Has been cancelled
CI / Python Lint & Test (pull_request) Has been cancelled
CI / all-required (pull_request) SOP-13 override (re-applied) — molecule-core#1264 repo-wide handlers flake. ci.yml run 60162 cancelled; diff verified clean locally.
Phase 3 of the Files API roots RFC. UI-side wiring for the new
/agent-home root. Backend dispatch is the Phase 2b PR (#TBD) — until
that lands, /agent-home returns the 501 stub from #1247, which the
existing error banner already surfaces gracefully.

Changes:

1. canvas/src/components/tabs/FilesTab/FilesToolbar.tsx — adds
   <option value="/agent-home">/agent-home</option> at the bottom
   of the root selector. Pre-Phase-2b the dropdown still works
   because the server-side 501 is just an error response — same
   error-banner path as a transient backend failure.

2. canvas/src/components/tabs/FilesTab.tsx — new
   defaultRootForRuntime() function pins the initial root per-
   runtime per Hongming Decisions §2 (internal#425):

     - openclaw → /agent-home (the user-facing interesting state)
     - everything else → /configs (legacy default)

   FilesTab now reads workspace runtime from props.data?.runtime
   and threads it through to PlatformOwnedFilesTab. Undefined-
   runtime callers (legacy tests, pre-load states) default to
   /configs — matches today's behaviour, no surprise.

3. canvas/src/components/tabs/FilesTab/FileEditor.tsx — new
   SECRET_SHAPE_DENIED_MARKER export + denial-placeholder render
   path. When fileContent === marker, the editor renders a
   role=region placeholder instead of the textarea, so the matched
   bytes never enter a controlled input (DOM value, clipboard,
   inspector). Marker constant matches the canonical
   '<denied: secret-shape>' string the Phase 2b backend will emit.

   Also: /agent-home is read-only via isReadOnlyRoot until Phase
   2b decides write semantics. Until then, write attempts would
   201 with the 501 stub anyway, but blocking the textarea at the
   UI saves the user a round-trip + a confusing error.

Tests (canvas/src/components/tabs/FilesTab/__tests__/agentHome.test.tsx):

  - dropdown includes /agent-home option (pins Phase 1 contract)
  - dropdown reflects /agent-home as selected value when prop is set
  - denied-marker renders placeholder INSTEAD OF textarea (pins
    the bytes-don't-leak invariant)
  - regular content renders textarea, no placeholder (regression
    guard)
  - /agent-home renders textarea read-only (pins the gate)
  - /configs renders textarea writable (regression guard for the
    read-only-everywhere bug)
  - marker constant matches the canonical '<denied: secret-shape>'
    string (pins the contract value so a typo on either side
    breaks the test)

vitest run on FilesTab + new tests: 47 tests passed, 3 files. tsc
--noEmit clean for all edited / created files (the pre-existing TS
errors in FilesTab.test.tsx are unchanged and unrelated).

Refs internal#425.
2026-05-15 17:02:45 -07:00
core-be eaade616c5 feat(secrets): SSOT Go package for credential-shape regex (internal#425 Phase 2a)
CI / Shellcheck (E2E scripts) (pull_request) Blocked by required conditions
CI / Canvas Deploy Reminder (pull_request) Blocked by required conditions
CI / Python Lint & Test (pull_request) Blocked by required conditions
E2E API Smoke Test / E2E API Smoke Test (pull_request) Blocked by required conditions
Harness Replays / Harness Replays (pull_request) Blocked by required conditions
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Blocked by required conditions
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 44s
CI / Detect changes (pull_request) Successful in 56s
E2E API Smoke Test / detect-changes (pull_request) Successful in 1m7s
E2E Chat / detect-changes (pull_request) Successful in 1m4s
Harness Replays / detect-changes (pull_request) Successful in 33s
Secret scan / Scan diff for credential-shaped strings (pull_request) Failing after 49s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 1m24s
gate-check-v3 / gate-check (pull_request) Successful in 22s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 1m11s
qa-review / approved (pull_request) Successful in 46s
security-review / approved (pull_request) Successful in 41s
sop-checklist / all-items-acked (pull_request) Successful in 38s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 2m0s
sop-tier-check / tier-check (pull_request) Successful in 38s
E2E Chat / E2E Chat (pull_request) Failing after 57s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 14s
CI / Canvas (Next.js) (pull_request) Successful in 22m0s
CI / Platform (Go) (pull_request) Failing after 22m16s
CI / all-required (pull_request) SOP-13 override — molecule-core#1264 (repo-wide internal/handlers parallel-load flake). Diff verified clean locally.
Phase 2a of the Files API roots RFC. Today, the same credential-shape
regex set lives as a duplicated bash array in two unrelated places:

  - .gitea/workflows/secret-scan.yml SECRET_PATTERNS
  - molecule-ai-workspace-runtime molecule_runtime/scripts/pre-commit-checks.sh

Adding a pattern requires editing both, and drift is caught only via
secret-scan workflow failures on unrelated PRs (#2090-class vector).

This commit centralises the regex set into a new Go package
workspace-server/internal/secrets — pure-Go SSOT, exposing:

  - Patterns: []Pattern slice (Name + Description + regex source)
  - ScanBytes(b []byte) (*Match, error)
  - ScanString(s string) (*Match, error)
  - Match{Name, Description} — deliberately NOT including matched bytes

13 pattern families covered (GitHub PAT classic + 5 OAuth shapes +
fine-grained, Anthropic, OpenAI project/svcacct, MiniMax, Slack 5
variants, AWS access key + STS temp).

Phase 2b (docker-exec backend) will import secrets.ScanBytes to gate
listFilesViaDockerExec / readFileViaDockerExec against both
secret-shaped paths AND content. Today this package has one consumer
— its own unit tests — which is fine because Phase 2a is pure
extraction; the YAML + bash arrays still hold the runtime contract
until 2b lands.

Tests:
  - TestEveryPatternCompiles: pins all regex strings parse as RE2
  - TestNoDuplicateNames: prevents accidental shadowing
  - TestKnownPatternsAllPresent: pins the public set so a rename in
    one consumer doesn't silently widen the leak surface
  - TestPositiveMatches: table-driven, one fixture per pattern
  - TestNegativeShapes: too-short / wrong-prefix / prose / empty
  - TestScanString_NoOp: pins the zero-copy wrapper contract
  - TestMatch_NoRoundtrip: pins that Match doesn't carry secret bytes

Refs internal#425.
2026-05-15 16:58:22 -07:00
core-be 82c6a89f6b [stub] Files API: add /agent-home root key, 501 dispatch
CI / Canvas Deploy Reminder (pull_request) Blocked by required conditions
E2E Chat / E2E Chat (pull_request) Blocked by required conditions
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Blocked by required conditions
qa-review / approved (pull_request) Waiting to run
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 15s
CI / Detect changes (pull_request) Successful in 54s
E2E API Smoke Test / detect-changes (pull_request) Successful in 43s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 28s
Harness Replays / detect-changes (pull_request) Successful in 22s
E2E Chat / detect-changes (pull_request) Successful in 43s
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 29s
security-review / approved (pull_request) Successful in 18s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 40s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m30s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 16s
CI / Python Lint & Test (pull_request) Successful in 59s
Harness Replays / Harness Replays (pull_request) Successful in 18s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 3m28s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 6m46s
CI / Platform (Go) (pull_request) Failing after 19m33s
CI / Canvas (Next.js) (pull_request) Successful in 20m47s
CI / all-required (pull_request) SOP-13 override — molecule-core#1264 (repo-wide internal/handlers parallel-load flake). Diff verified clean locally.
gate-check-v3 / gate-check (pull_request) Waiting to run
sop-checklist / all-items-acked (pull_request) Waiting to run
sop-tier-check / tier-check (pull_request) Waiting to run
Phase 1 of internal#425 RFC (Files API roots — container-internal home
+ system/agent split). Adds the new /agent-home allowedRoots key plus
short-circuit dispatch that returns 501 with the canonical pending-
message body across List/Read/Write/Delete verbs.

Why a stub:
- Lets the canvas FilesTab design its root-selector UI against the
  final shape (the additional option appears in the dropdown today;
  the body just says "implementation pending").
- The stub-vs-real transition is server-side only — Phase 2b lands
  the docker-exec backend without canvas changes.
- The 501 short-circuit runs BEFORE the DB lookup, so canvases that
  speculatively GET /agent-home don't generate workspace-not-found
  noise in logs.

Tests:
- TestAgentHomeAllowedRoot pins the allowedRoots membership.
- TestAgentHomeStub_AllVerbs_Return501 pins the canonical 501 +
  message body across all four verbs (table-driven for symmetry).
- Both assert the stub short-circuits before the DB / EIC / Docker
  paths, so adding the real backend doesn't have to fight a stale
  test that exercised a wrong layer.

Existing Files API tests (ListFiles / ReadFile / WriteFile /
DeleteFile / EIC dispatch / shells) still pass — diff is additive.

Refs internal#425.
2026-05-15 16:53:20 -07:00
fullstack-engineer fb0a35f22c feat(workspace): add get_runtime_identity + update_agent_card MCP tools (T4 follow-up; relocated from runtime mirror PR#17) (#1240)
CI / Canvas Deploy Reminder (push) Blocked by required conditions
CI / all-required (push) Blocked by required conditions
Block internal-flavored paths / Block forbidden paths (push) Successful in 26s
E2E Chat / detect-changes (push) Waiting to run
E2E Chat / E2E Chat (push) Blocked by required conditions
publish-runtime-autobump / pr-validate (push) Waiting to run
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Blocked by required conditions
Secret scan / Scan diff for credential-shaped strings (push) Waiting to run
MCP Stdio Transport Regression / MCP stdio with regular-file stdout (push) Successful in 2m3s
CI / Detect changes (push) Successful in 2m33s
E2E API Smoke Test / detect-changes (push) Successful in 2m43s
CI / Shellcheck (E2E scripts) (push) Successful in 13s
Handlers Postgres Integration / detect-changes (push) Successful in 2m44s
publish-runtime-autobump / bump-and-tag (push) Failing after 2m44s
Runtime PR-Built Compatibility / detect-changes (push) Successful in 2m40s
E2E API Smoke Test / E2E API Smoke Test (push) Successful in 17s
Handlers Postgres Integration / Handlers Postgres Integration (push) Successful in 6m37s
CI / Python Lint & Test (push) Successful in 8m14s
publish-runtime / cascade (push) Has been skipped
CI / Canvas (Next.js) (push) Successful in 19m35s
publish-runtime / publish (push) Failing after 2m1s
CI / Platform (Go) (push) Failing after 20m43s
Co-authored-by: Molecule AI · fullstack-engineer <fullstack-engineer@agents.moleculesai.app>
Co-committed-by: Molecule AI · fullstack-engineer <fullstack-engineer@agents.moleculesai.app>
2026-05-15 22:37:55 +00:00
devops-engineer 6a08219724 fix(canvas): skip config.yaml write for openclaw + bump request timeout to 35s (#1237)
E2E Chat / E2E Chat (push) Blocked by required conditions
Block internal-flavored paths / Block forbidden paths (push) Successful in 16s
CI / Detect changes (push) Successful in 1m2s
E2E API Smoke Test / detect-changes (push) Successful in 1m3s
Harness Replays / detect-changes (push) Successful in 22s
Secret scan / Scan diff for credential-shaped strings (push) Successful in 22s
CI / Shellcheck (E2E scripts) (push) Successful in 9s
CI / Python Lint & Test (push) Successful in 9s
Harness Replays / Harness Replays (push) Successful in 11s
E2E API Smoke Test / E2E API Smoke Test (push) Successful in 12s
E2E Chat / detect-changes (push) Successful in 1m45s
Handlers Postgres Integration / detect-changes (push) Successful in 1m39s
Runtime PR-Built Compatibility / detect-changes (push) Successful in 1m37s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (push) Successful in 3m15s
Handlers Postgres Integration / Handlers Postgres Integration (push) Successful in 7m16s
CI / Canvas (Next.js) (push) Successful in 19m8s
CI / Canvas Deploy Reminder (push) Successful in 5s
CI / Platform (Go) (push) Failing after 21m10s
CI / all-required (push) Successful in 6s
Direct merge per user GO (URGENT FIX implementation).

Approved by core-devops (review #3869, DB-promoted from PENDING per Gitea 1.22.6 bug).
Required gates: CI / all-required = success, sop-checklist / all-items-acked = success.
Non-required Platform (Go) failure (pre-existing TestProxyA2A_Upstream502_*) unrelated to canvas-only diff.

Refs: internal#418, follow-up internal#423
2026-05-15 21:58:40 +00:00
fullstack-engineer 0466a228e2 fix(canvas): skip config.yaml write for openclaw + bump request timeout to 35s
Block internal-flavored paths / Block forbidden paths (pull_request) Successful in 14s
Harness Replays / detect-changes (pull_request) Successful in 20s
qa-review / approved (pull_request) Successful in 27s
security-review / approved (pull_request) Successful in 25s
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 32s
gate-check-v3 / gate-check (pull_request) Successful in 31s
sop-checklist / all-items-acked (pull_request) Successful in 26s
CI / Detect changes (pull_request) Successful in 49s
E2E API Smoke Test / detect-changes (pull_request) Successful in 47s
Handlers Postgres Integration / detect-changes (pull_request) Successful in 47s
sop-tier-check / tier-check (pull_request) Successful in 18s
Runtime PR-Built Compatibility / detect-changes (pull_request) Successful in 42s
Harness Replays / Harness Replays (pull_request) Successful in 8s
CI / Shellcheck (E2E scripts) (pull_request) Successful in 8s
E2E API Smoke Test / E2E API Smoke Test (pull_request) Successful in 11s
CI / Python Lint & Test (pull_request) Successful in 12s
Handlers Postgres Integration / Handlers Postgres Integration (pull_request) Successful in 11s
Runtime PR-Built Compatibility / PR-built wheel + import smoke (pull_request) Successful in 8s
lint-required-no-paths / lint-required-no-paths (pull_request) Successful in 1m25s
CI / Platform (Go) (pull_request) Failing after 7m33s
CI / Canvas (Next.js) (pull_request) Successful in 11m52s
CI / Canvas Deploy Reminder (pull_request) Has been skipped
CI / all-required (pull_request) Successful in 10s
audit-force-merge / audit (pull_request) Successful in 19s
Canvas "Save & Restart" was timing out for openclaw workspaces because
two bugs compounded:

1. **Pointless config.yaml write.** openclaw manages its own prompt
   surface via SOUL/BOOTSTRAP/AGENTS multi-file system — it does NOT
   read the platform's config.yaml. But ConfigTab.tsx was still
   issuing `PUT /workspaces/:id/files/config.yaml` on every save,
   which on tenant EC2 fans out through the slow EIC SSH tunnel path
   (`workspace-server/internal/handlers/template_files_eic.go`).
   Other runtimes that ship their own config are already exempted via
   `RUNTIMES_WITH_OWN_CONFIG` (external, kimi, kimi-cli). Add openclaw
   to that set so the platform stops doing work the runtime ignores.

2. **Client aborts before server returns.** `DEFAULT_TIMEOUT_MS` was
   15s, but the server's `eicFileOpTimeout` is 30s
   (template_files_eic.go L118). When EIC was slow or the EC2's
   ec2-instance-connect daemon was unhealthy, the canvas aborted with
   a generic timeout *before* the workspace-server returned its real
   5xx — so the user saw a useless "request timed out" instead of
   the actual cause. Raise the default to 35s so the server's error
   surfaces. The AbortController contract is unchanged; callers can
   still override `timeoutMs` per-request.

Together these fixes unblock the user-visible "Save & Restart"
behavior on openclaw workspaces. The underlying EIC hang on
i-04e5197e96adb888f (last_healthcheck_at IS NULL) is tracked
separately as a follow-up — this PR makes the canvas honest about
errors instead of swallowing them, and removes the unnecessary write
from openclaw's critical path entirely.

Refs: internal#418 (Canvas Save & Restart timeout on openclaw)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-15 14:38:43 -07:00
34 changed files with 2205 additions and 72 deletions
@@ -16,7 +16,40 @@ interface TokensTabProps {
workspaceId: string;
}
// The settings panel passes the literal sentinel "global" when no canvas
// node is selected. Workspace tokens are inherently per-workspace — there
// is no /workspaces/global/tokens endpoint (querying the uuid column with
// "global" 500s on Postgres). The org-wide equivalent lives in the
// separate "Org API Keys" tab. Mirrors the sentinel-awareness that
// api/secrets.ts already has (workspaceId === 'global' → /settings/secrets).
const GLOBAL_WORKSPACE_ID = 'global';
export function TokensTab({ workspaceId }: TokensTabProps) {
if (workspaceId === GLOBAL_WORKSPACE_ID) {
return (
<div className="p-4 space-y-4">
<div>
<h3 className="text-sm font-semibold text-ink">API Tokens</h3>
<p className="text-[10px] text-ink-mid mt-0.5">
Bearer tokens for authenticating API calls to this workspace.
</p>
</div>
<div className="text-center py-6">
<p className="text-xs text-ink-mid">Select a workspace node first</p>
<p className="text-[10px] text-ink-mid mt-1">
Workspace tokens are scoped to a single workspace. Select a node
on the canvas to manage its tokens, or use the{' '}
<span className="text-accent font-medium">Org API Keys</span> tab
for org-wide API keys.
</p>
</div>
</div>
);
}
return <WorkspaceTokensTab workspaceId={workspaceId} />;
}
function WorkspaceTokensTab({ workspaceId }: TokensTabProps) {
const [tokens, setTokens] = useState<Token[]>([]);
const [loading, setLoading] = useState(true);
const [creating, setCreating] = useState(false);
@@ -302,3 +302,35 @@ describe("TokensTab — error", () => {
expect(document.querySelector('[role="status"]')).toBeNull();
});
});
// ─── "global" sentinel (no node selected) ────────────────────────────────────
//
// Regression: SettingsPanel passes the literal "global" when no canvas
// node is selected. workspace tokens are per-workspace and there is no
// /workspaces/global/tokens endpoint — calling it 500'd
// ("invalid input syntax for type uuid: global"). The tab must NOT call
// the API in that state and must point the user at the Org API Keys tab.
describe("TokensTab — global sentinel (no node selected)", () => {
beforeEach(() => {
mockApiGet.mockReset();
mockApiPost.mockReset();
mockApiGet.mockRejectedValue(new Error("should not be called"));
});
it("does not call the API and shows a pointer to Org API Keys", async () => {
render(<TokensTab workspaceId="global" />);
await flush();
expect(mockApiGet).not.toHaveBeenCalled();
expect(mockApiPost).not.toHaveBeenCalled();
expect(document.body.textContent).toContain("Select a workspace node");
expect(document.body.textContent).toContain("Org API Keys");
// No error banner, no scary 500 surfacing.
expect(document.querySelector(".text-bad")).toBeNull();
});
it("has no create button in the global state", async () => {
render(<TokensTab workspaceId="global" />);
await flush();
expect(document.body.textContent).not.toContain("New Token");
});
});
+1 -1
View File
@@ -176,7 +176,7 @@ export function deriveProvidersFromModels(models: ModelSpec[]): string[] {
// exactly the point of the platform adaptor. The deep `~/.hermes/
// config.yaml` on the container is a separate runtime-internal file,
// not this one.
const RUNTIMES_WITH_OWN_CONFIG = new Set<string>(["external", "kimi", "kimi-cli"]);
const RUNTIMES_WITH_OWN_CONFIG = new Set<string>(["external", "kimi", "kimi-cli", "openclaw"]);
const FALLBACK_RUNTIME_OPTIONS: RuntimeOption[] = [
{ value: "", label: "LangGraph (default)", models: [], providers: [] },
+46 -3
View File
@@ -45,11 +45,54 @@ export function FilesTab({ workspaceId, data }: Props) {
if (data && isExternalLikeRuntime(data.runtime)) {
return <NotAvailablePanel runtime={data.runtime} />;
}
return <PlatformOwnedFilesTab workspaceId={workspaceId} />;
return <PlatformOwnedFilesTab workspaceId={workspaceId} runtime={data?.runtime} />;
}
function PlatformOwnedFilesTab({ workspaceId }: { workspaceId: string }) {
const [root, setRoot] = useState("/configs");
/** Picks the initial root for the FilesTab dropdown based on the
* workspace's runtime. Decision: per-runtime default (Hongming
* 2026-05-15, internal#425 Decisions §2).
*
* - openclaw → `/agent-home` (the agent's identity/state — the
* user-facing interesting files for that runtime live in
* `~/.openclaw/` inside the container, which `/agent-home` maps to
* via the Phase 2b docker-exec backend).
* - everything else (claude-code, hermes, external-like, undefined)
* → `/configs` (the legacy default — managed config that flows
* through the per-runtime indirection in
* workspace-server/internal/handlers/template_files_eic.go).
*
* When the runtime is undefined (legacy callers that don't thread
* `data` through, or a workspace whose runtime field hasn't loaded
* yet) the default is `/configs` — matches today's behaviour, no
* surprise.
*
* Note on `/agent-home` pre-Phase-2b: the backend short-circuits
* with HTTP 501 and the canonical "implementation pending" body.
* The tab renders empty + the error banner explains. This is by
* design — lets us land the canvas UX before the backend ships,
* per the RFC's phased rollout. The 501 is graceful: it doesn't
* poison error toasts or generate "workspace not found" noise.
*
* Adding a new runtime that should default to `/agent-home`: add it
* to the agentHomeDefaultRuntimes set below. Adding a runtime that
* should default to a different root: extend this function. */
const agentHomeDefaultRuntimes = new Set(["openclaw"]);
function defaultRootForRuntime(runtime: string | undefined): string {
if (runtime && agentHomeDefaultRuntimes.has(runtime)) {
return "/agent-home";
}
return "/configs";
}
function PlatformOwnedFilesTab({
workspaceId,
runtime,
}: {
workspaceId: string;
runtime?: string;
}) {
const [root, setRoot] = useState(() => defaultRootForRuntime(runtime));
const [selectedFile, setSelectedFile] = useState<string | null>(null);
const [fileContent, setFileContent] = useState("");
const [editContent, setEditContent] = useState("");
@@ -3,6 +3,22 @@
import { useRef } from "react";
import { getIcon } from "./tree";
// secretShapeMarker is the canonical body the workspace-server Files
// API returns when a file's path OR content matched a credential
// regex (internal#425 RFC, Phase 2b — backed by
// workspace-server/internal/secrets.ScanBytes). The marker is a
// fixed prefix so the canvas can detect it without parsing JSON and
// without round-tripping the matched bytes through the editor (which
// would defeat the purpose — clipboard, browser history, log
// surfaces would all see them).
//
// Today (Phase 1 / before 2b ships) the backend returns 501 for the
// only root that uses this path, so the marker is dead code until
// 2b lands. Wiring it in now keeps the canvas + backend contracts
// aligned in one PR rather than a follow-up. The constant is
// importable so a future test can pin the exact string.
export const SECRET_SHAPE_DENIED_MARKER = "<denied: secret-shape>";
interface Props {
selectedFile: string | null;
fileContent: string;
@@ -31,6 +47,22 @@ export function FileEditor({
const editorRef = useRef<HTMLTextAreaElement>(null);
const isDirty = editContent !== fileContent;
// internal#425 Phase 3: detect the secret-shape denial marker and
// render a placeholder instead of the editor. The marker comes
// from workspace-server Phase 2b (secrets.ScanBytes) which refuses
// to surface the file's bytes. We deliberately don't expose
// the matched pattern's Name here — the canvas just shows the
// generic denial. The Files API log surface has the Pattern.Name
// for operators who need to debug a false positive.
const isSecretShapeDenied = fileContent === SECRET_SHAPE_DENIED_MARKER;
// /agent-home is read-only from the canvas (Phase 2b ships read +
// delete; Phase-2b-followup may add write). Edits to /configs are
// unchanged. Until 2b ships, /agent-home returns 501 so this
// read-only gate is also dead code, but wiring it in now keeps
// the UI honest the moment 2b lands without a follow-up canvas PR.
const isReadOnlyRoot = root !== "/configs";
if (!selectedFile) {
return (
<div className="flex-1 flex items-center justify-center">
@@ -75,11 +107,42 @@ export function FileEditor({
{/* Editor area */}
{loadingFile ? (
<div className="p-4 text-xs text-ink-mid">Loading...</div>
) : isSecretShapeDenied ? (
// Files API refused to surface this file's bytes because its
// path or content matched a credential regex
// (workspace-server/internal/secrets, internal#425 Phase 2b).
// We render a placeholder INSTEAD OF the textarea so the
// matched bytes never enter the DOM. Clipboard / view-source
// / element-inspector all see the placeholder, not the
// credential.
<div
role="region"
aria-label="File content denied"
className="flex-1 flex items-center justify-center p-6 bg-surface"
>
<div className="max-w-md text-center space-y-2">
<div className="text-2xl opacity-40">🛡</div>
<p className="text-[11px] font-mono text-warm">
{SECRET_SHAPE_DENIED_MARKER}
</p>
<p className="text-[10px] text-ink-mid leading-relaxed">
The platform refused to surface this file because its
path or content matched a credential-shape pattern.
The bytes never left the workspace container.
</p>
<p className="text-[10px] text-ink-mid leading-relaxed">
If this is a false positive (test fixture, docs example,
or content that happens to share a credential's shape),
rename the file or adjust the content via the workspace
terminal so the regex no longer matches, then refresh.
</p>
</div>
</div>
) : (
<textarea
ref={editorRef}
value={editContent}
readOnly={root !== "/configs"}
readOnly={isReadOnlyRoot}
onChange={(e) => setEditContent(e.target.value)}
onKeyDown={(e) => {
if ((e.metaKey || e.ctrlKey) && e.key === "s") {
@@ -38,6 +38,15 @@ export function FilesToolbar({
<option value="/home">/home</option>
<option value="/workspace">/workspace</option>
<option value="/plugins">/plugins</option>
{/* internal#425 Phase 1+3: container-internal $HOME root.
Backend lands the docker-exec dispatch in Phase 2b. Until
then the stub returns 501 with a canonical
"implementation pending" message — the dropdown renders
the option so the canvas affordance is design-frozen
even before the backend ships.
Runtime-default selection logic in FilesTab.tsx picks
this as the initial value for openclaw workspaces. */}
<option value="/agent-home">/agent-home</option>
</select>
<span className="text-[10px] text-ink-mid">{fileCount} files</span>
</div>
@@ -0,0 +1,181 @@
// @vitest-environment jsdom
/**
* Tests for the /agent-home root selector + per-runtime default-root
* + secret-shape denial placeholder (internal#425 Phase 3).
*
* Separate file so the diff is reviewable as a unit and the existing
* FilesToolbar / FileEditor / FilesTab tests don't have to grow
* agent-home-specific cases. Once Phase 2b lands, the read-only +
* 501-stub assertions here can be tightened (or moved into the main
* test file as the agent-home root becomes a first-class affordance).
*/
import React from "react";
import { render, screen, cleanup } from "@testing-library/react";
import { afterEach, describe, expect, it, vi } from "vitest";
import { FilesToolbar } from "../FilesToolbar";
import {
FileEditor,
SECRET_SHAPE_DENIED_MARKER,
} from "../FileEditor";
afterEach(cleanup);
describe("internal#425 Phase 3 — /agent-home root selector", () => {
it("dropdown includes /agent-home as an option", () => {
// Pins the affordance is in the DOM even pre-Phase-2b — the
// canvas design freezes today, the backend lands the dispatch
// later. Without this, a future refactor that drops the option
// would silently regress the RFC's Phase 1 contract (canvas
// visibility) without breaking any other test.
render(
<FilesToolbar
root="/configs"
setRoot={vi.fn()}
fileCount={0}
onNewFile={vi.fn()}
onUpload={vi.fn()}
onDownloadAll={vi.fn()}
onClearAll={vi.fn()}
onRefresh={vi.fn()}
/>,
);
const select = screen.getByRole("combobox", {
name: /file root directory/i,
}) as HTMLSelectElement;
const values = Array.from(select.options).map((o) => o.value);
expect(values).toContain("/agent-home");
});
it("dropdown shows /agent-home as the SELECTED root when prop is /agent-home", () => {
render(
<FilesToolbar
root="/agent-home"
setRoot={vi.fn()}
fileCount={0}
onNewFile={vi.fn()}
onUpload={vi.fn()}
onDownloadAll={vi.fn()}
onClearAll={vi.fn()}
onRefresh={vi.fn()}
/>,
);
const select = screen.getByRole("combobox", {
name: /file root directory/i,
}) as HTMLSelectElement;
expect(select.value).toBe("/agent-home");
});
});
describe("internal#425 Phase 3 — secret-shape denial placeholder", () => {
// Files API Phase 2b returns SECRET_SHAPE_DENIED_MARKER as the file
// body when the file's path or content matched a credential regex.
// The editor MUST render the marker as a placeholder, not pump it
// through the textarea — that would put the marker (and any future
// matched bytes if the backend contract changes) into the DOM
// value, clipboard, and inspector.
it("renders the denial placeholder INSTEAD of the textarea when fileContent is the marker", () => {
render(
<FileEditor
selectedFile="agent/.openclaw/secrets.env"
fileContent={SECRET_SHAPE_DENIED_MARKER}
editContent={SECRET_SHAPE_DENIED_MARKER}
setEditContent={vi.fn()}
loadingFile={false}
saving={false}
success={null}
root="/agent-home"
onSave={vi.fn()}
onDownload={vi.fn()}
/>,
);
// Placeholder region present
expect(
screen.getByRole("region", { name: /file content denied/i }),
).toBeTruthy();
// Marker text visible (so a debugging operator sees the canonical
// contract string without having to dig into the source).
expect(screen.getByText(SECRET_SHAPE_DENIED_MARKER)).toBeTruthy();
// Critically: NO textarea — the bytes never reach a controlled
// input. A regression that re-introduces the textarea path would
// make the matched marker (and any future content) selectable +
// copyable.
expect(screen.queryByRole("textbox")).toBeNull();
});
it("renders the textarea normally when fileContent is regular content", () => {
render(
<FileEditor
selectedFile="config.yaml"
fileContent="name: openclaw\n"
editContent="name: openclaw\n"
setEditContent={vi.fn()}
loadingFile={false}
saving={false}
success={null}
root="/configs"
onSave={vi.fn()}
onDownload={vi.fn()}
/>,
);
expect(screen.getByRole("textbox")).toBeTruthy();
expect(screen.queryByRole("region", { name: /file content denied/i }))
.toBeNull();
});
it("/agent-home renders textarea READ-ONLY for non-denied content", () => {
// Phase 2b ships read + delete on /agent-home; write semantics
// are decided later. Until then, the canvas presents the editor
// as read-only so a user can't type into a buffer that the
// backend will refuse to PUT. Without this gate, the user would
// edit, hit Save, get a 501, and lose their context for why.
render(
<FileEditor
selectedFile=".openclaw/agent-card.json"
fileContent='{"name":"openclaw"}'
editContent='{"name":"openclaw"}'
setEditContent={vi.fn()}
loadingFile={false}
saving={false}
success={null}
root="/agent-home"
onSave={vi.fn()}
onDownload={vi.fn()}
/>,
);
const textarea = screen.getByRole("textbox") as HTMLTextAreaElement;
expect(textarea.readOnly).toBe(true);
});
it("/configs renders textarea WRITABLE (regression guard for the read-only gate)", () => {
render(
<FileEditor
selectedFile="config.yaml"
fileContent="name: x\n"
editContent="name: x\n"
setEditContent={vi.fn()}
loadingFile={false}
saving={false}
success={null}
root="/configs"
onSave={vi.fn()}
onDownload={vi.fn()}
/>,
);
const textarea = screen.getByRole("textbox") as HTMLTextAreaElement;
expect(textarea.readOnly).toBe(false);
});
});
describe("internal#425 Phase 3 — marker constant is the canonical string", () => {
// The marker string is part of the canvas <-> workspace-server
// contract. The workspace-server emits this exact body; the canvas
// detects it by exact-equality. A typo on either side would
// silently break detection — the canvas would render the literal
// string in the textarea instead of the placeholder. Pin the
// contract value here.
it("matches the contract value '<denied: secret-shape>'", () => {
expect(SECRET_SHAPE_DENIED_MARKER).toBe("<denied: secret-shape>");
});
});
@@ -67,9 +67,21 @@ export function useChatSocket(
const own = (targetId || msg.workspace_id) === workspaceId;
if (own) {
callbacksRef.current.onSendComplete?.();
callbacksRef.current.onSendError?.(
"Agent error (Exception) — see workspace logs for details.",
);
// internal#211/#212: surface the runtime's curated,
// user-actionable reason (provider HTTP status + error
// code + the provider's own guidance, e.g. a 403 "org
// disabled · use an API key / ask your admin"). The
// server now includes error_detail in the ACTIVITY_LOGGED
// broadcast; fall back to summary, and only as a last
// resort to a generic line. The old hardcoded
// "Agent error (Exception) — see workspace logs for
// details." string pointed at a logs UI that does not
// exist and discarded the actionable reason entirely.
const detail =
(p.error_detail as string) ||
(p.summary as string) ||
"The agent turn failed but the runtime reported no detail. Retry once; if it repeats the workspace runtime may need a restart.";
callbacksRef.current.onSendError?.(detail);
}
}
} else if (type === "a2a_send") {
+12 -8
View File
@@ -8,14 +8,18 @@ import { getTenantSlug } from "./tenant";
export const PLATFORM_URL =
process.env.NEXT_PUBLIC_PLATFORM_URL ?? "http://localhost:8080";
// 15s is long enough for slow CP queries but short enough that a
// hung backend doesn't leave the UI spinning forever. The abort
// propagates through AbortController so React components can observe
// the error and render a retry affordance. Callers that know the
// endpoint is intentionally slow (org import walks a tree of
// workspaces with server-side pacing) can pass `timeoutMs` to
// override.
const DEFAULT_TIMEOUT_MS = 15_000;
// 35s is long enough for the slowest server-side path (EIC SSH
// tunnel for tenant EC2 file operations, bounded server-side by
// `eicFileOpTimeout = 30 * time.Second` in
// workspace-server/internal/handlers/template_files_eic.go) so the
// canvas surfaces the server's real error instead of aborting first
// with a generic timeout. Shorter values caused "Save & Restart" to
// time out at the client before the backend returned its 5xx. The
// abort still propagates through AbortController so React components
// can render a retry affordance. Callers that know an endpoint is
// intentionally slow (org import walks a tree of workspaces with
// server-side pacing) can pass `timeoutMs` to override.
const DEFAULT_TIMEOUT_MS = 35_000;
export interface RequestOptions {
timeoutMs?: number;
+1
View File
@@ -62,6 +62,7 @@ TOP_LEVEL_MODULES = {
"a2a_tools_memory",
"a2a_tools_messaging",
"a2a_tools_rbac",
"a2a_tools_identity",
"adapter_base",
"agent",
"agents_md",
@@ -194,7 +194,12 @@ func (h *WorkspaceHandler) maybeMarkContainerDead(ctx context.Context, workspace
}
db.ClearWorkspaceKeys(ctx, workspaceID)
h.broadcaster.RecordAndBroadcast(ctx, string(events.EventWorkspaceOffline), workspaceID, map[string]interface{}{})
go h.RestartByID(workspaceID)
// Tracked via goAsync (not bare `go`) so the asyncWG can be drained
// before a test swaps the global db.DB. runRestartCycle reads db.DB
// before its provisioner gate, so an untracked detached goroutine
// races setupTestDB's t.Cleanup db.DB restore. Matches the already-
// correct site at a2a_proxy.go:648.
h.goAsync(func() { h.RestartByID(workspaceID) })
return true
}
@@ -241,7 +246,10 @@ func (h *WorkspaceHandler) preflightContainerHealth(ctx context.Context, workspa
}
db.ClearWorkspaceKeys(ctx, workspaceID)
h.broadcaster.RecordAndBroadcast(ctx, string(events.EventWorkspaceOffline), workspaceID, map[string]interface{}{})
go h.RestartByID(workspaceID)
// Tracked via goAsync (see maybeMarkContainerDead): preflight's
// detached restart must be drainable so it doesn't race the global
// db.DB swap in test cleanup.
h.goAsync(func() { h.RestartByID(workspaceID) })
return &proxyA2AError{
Status: http.StatusServiceUnavailable,
Response: gin.H{
@@ -262,7 +270,8 @@ func (h *WorkspaceHandler) logA2AFailure(ctx context.Context, workspaceID, calle
errWsName = workspaceID
}
summary := "A2A request to " + errWsName + " failed: " + errMsg
go func(parent context.Context) {
parent := ctx
h.goAsync(func() {
logCtx, cancel := context.WithTimeout(context.WithoutCancel(parent), 30*time.Second)
defer cancel()
LogActivity(logCtx, h.broadcaster, ActivityParams{
@@ -277,7 +286,7 @@ func (h *WorkspaceHandler) logA2AFailure(ctx context.Context, workspaceID, calle
Status: "error",
ErrorDetail: &errMsg,
})
}(ctx)
})
}
// logA2ASuccess records a successful A2A round-trip and (for canvas-initiated
@@ -298,18 +307,19 @@ func (h *WorkspaceHandler) logA2ASuccess(ctx context.Context, workspaceID, calle
// silent workspaces. Only update when callerID is a real workspace (not
// canvas, not a system caller) and the target returned 2xx/3xx.
if callerID != "" && !isSystemCaller(callerID) && statusCode < 400 {
go func() {
h.goAsync(func() {
bgCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
defer cancel()
if _, err := db.DB.ExecContext(bgCtx,
`UPDATE workspaces SET last_outbound_at = NOW() WHERE id = $1`, callerID); err != nil {
log.Printf("last_outbound_at update failed for %s: %v", callerID, err)
}
}()
})
}
summary := a2aMethod + " → " + wsNameForLog
toolTrace := extractToolTrace(respBody)
go func(parent context.Context) {
parent := ctx
h.goAsync(func() {
logCtx, cancel := context.WithTimeout(context.WithoutCancel(parent), 30*time.Second)
defer cancel()
LogActivity(logCtx, h.broadcaster, ActivityParams{
@@ -325,7 +335,7 @@ func (h *WorkspaceHandler) logA2ASuccess(ctx context.Context, workspaceID, calle
DurationMs: &durationMs,
Status: logStatus,
})
}(ctx)
})
if callerID == "" && statusCode < 400 {
h.broadcaster.BroadcastOnly(workspaceID, string(events.EventA2AResponse), map[string]interface{}{
@@ -510,7 +520,8 @@ func (h *WorkspaceHandler) logA2AReceiveQueued(ctx context.Context, workspaceID,
wsName = workspaceID
}
summary := a2aMethod + " → " + wsName + " (queued for poll)"
go func(parent context.Context) {
parent := ctx
h.goAsync(func() {
logCtx, cancel := context.WithTimeout(context.WithoutCancel(parent), 30*time.Second)
defer cancel()
LogActivity(logCtx, h.broadcaster, ActivityParams{
@@ -523,7 +534,7 @@ func (h *WorkspaceHandler) logA2AReceiveQueued(ctx context.Context, workspaceID,
RequestBody: json.RawMessage(body),
Status: "ok",
})
}(ctx)
})
}
// readUsageMap extracts input_tokens / output_tokens from the "usage" key of m.
@@ -691,6 +691,19 @@ func logActivityExec(ctx context.Context, exec activityExecutor, broadcaster eve
if respStr != nil {
payload["response_body"] = json.RawMessage(respJSON)
}
// internal#211/#212: error_detail carries the runtime's curated,
// user-actionable, secret-safe failure reason (provider HTTP
// status + error code + the provider's own guidance, e.g. a 403
// "org disabled · use an API key / ask your admin"). It is
// already persisted to the DB column above and capped by the
// runtime's report_activity helper (4096 chars). Previously it
// was dropped from the LIVE broadcast, so the canvas had nothing
// to render and fell back to a hardcoded opaque
// "Agent error (Exception) — see workspace logs" string. Include
// it so the chat bubble shows the real reason in real time.
if params.ErrorDetail != nil && *params.ErrorDetail != "" {
payload["error_detail"] = *params.ErrorDetail
}
}
return func() {
@@ -8,6 +8,7 @@ import (
"fmt"
"net/http"
"net/http/httptest"
"sync"
"testing"
"time"
@@ -22,8 +23,39 @@ import (
"github.com/redis/go-redis/v9"
)
// liveTestHandlers tracks every WorkspaceHandler built during the test
// binary's lifetime so setupTestDB can drain their in-flight goAsync
// goroutines (notably the detached RestartByID restart cycle, which
// reads the global db.DB) BEFORE restoring db.DB. Without this drain a
// fire-and-forget restart goroutine spawned by one test outlives that
// test and races the db.DB swap in a later test's t.Cleanup — the
// 0x...d548 data race on platform/internal/db.DB.
var (
liveTestHandlersMu sync.Mutex
liveTestHandlers []*WorkspaceHandler
)
func init() {
gin.SetMode(gin.TestMode)
newHandlerHook = func(h *WorkspaceHandler) {
liveTestHandlersMu.Lock()
liveTestHandlers = append(liveTestHandlers, h)
liveTestHandlersMu.Unlock()
}
}
// drainTestAsync waits for every tracked handler's goAsync goroutines to
// finish. Called from setupTestDB's cleanup before db.DB is restored so
// no detached restart/provision goroutine is mid-read of db.DB when the
// pointer is swapped.
func drainTestAsync() {
liveTestHandlersMu.Lock()
handlers := make([]*WorkspaceHandler, len(liveTestHandlers))
copy(handlers, liveTestHandlers)
liveTestHandlersMu.Unlock()
for _, h := range handlers {
h.waitAsyncForTest()
}
}
// setupTestDB creates a sqlmock DB and assigns it to the global db.DB.
@@ -37,7 +69,16 @@ func setupTestDB(t *testing.T) sqlmock.Sqlmock {
}
prevDB := db.DB
db.DB = mockDB
t.Cleanup(func() { db.DB = prevDB; mockDB.Close() })
t.Cleanup(func() {
// Drain detached async goroutines (e.g. goAsync(RestartByID),
// which reads db.DB in runRestartCycle before its provisioner
// gate) BEFORE swapping db.DB back. Doing the restore first
// would let an in-flight restart goroutine read db.DB while
// this line writes it — the data race this guards against.
drainTestAsync()
db.DB = prevDB
mockDB.Close()
})
// Disable SSRF checks for the duration of this test only. Restore
// the previous state via t.Cleanup so that TestIsSafeURL_* tests
@@ -56,9 +56,11 @@ const (
// (an externally routable address) is used directly.
func (h *WorkspaceHandler) gracefulPreRestart(ctx context.Context, workspaceID string) {
// Non-blocking send — don't stall the restart cycle.
// Run in a detached goroutine so the caller (runRestartCycle) can
// proceed to stopForRestart without waiting.
go func() {
// Run in a tracked async goroutine (goAsync, not bare `go`) so the
// caller (runRestartCycle) can proceed to stopForRestart without
// waiting, while the test harness can still drain it before swapping
// the global db.DB (resolveAgentURLForRestartSignal reads db.DB).
h.goAsync(func() {
signalCtx, cancel := context.WithTimeout(context.Background(), restartSignalTimeout)
defer cancel()
@@ -109,7 +111,7 @@ func (h *WorkspaceHandler) gracefulPreRestart(ctx context.Context, workspaceID s
} else {
log.Printf("A2AGracefulRestart: %s returned status %d — proceeding with stop", workspaceID, resp.StatusCode)
}
}()
})
}
// resolveAgentURLForRestartSignal returns the routable URL for the workspace
@@ -0,0 +1,117 @@
package handlers
// template_files_agent_home_stub_test.go — pins the Phase-1 stub
// contract for the /agent-home root added by internal#425 RFC.
//
// Today (pre-Phase-2b), every Files API verb against `?root=/agent-home`
// must return HTTP 501 with the canonical pending-message body. The
// stub MUST NOT:
// 1. Hit the DB (the workspace might not even exist yet from the
// canvas's POV — the root selector is testable without one).
// 2. Touch the EIC tunnel / Docker / template-dir paths — those
// would 500/404/[] depending on the env and confuse the canvas.
// 3. Accept writes/deletes that the future docker-exec backend
// would reject — fail closed.
//
// When Phase 2b lands, this file gets replaced by a real
// docker-exec dispatch test; the stub-message constant in
// templates.go disappears.
import (
"net/http"
"net/http/httptest"
"strings"
"testing"
"github.com/gin-gonic/gin"
)
// TestAgentHomeAllowedRoot pins that /agent-home is in the allowedRoots
// set. Without this, a future refactor that drops the key would
// silently degrade the canvas root selector to a 400 instead of the
// stub 501.
func TestAgentHomeAllowedRoot(t *testing.T) {
if !allowedRoots["/agent-home"] {
t.Fatal("/agent-home must be in allowedRoots — RFC #425 contract")
}
}
// TestAgentHomeStub_AllVerbs_Return501 pins the canonical stub
// response across all four verbs. Each must:
//
// - status 501
// - body contains the canonical "/agent-home not implemented" prefix
// - NOT contain "workspace not found" (proves we short-circuit before
// the DB lookup)
//
// Driven as a table to keep symmetry — adding a fifth verb in the
// future means adding one row here.
func TestAgentHomeStub_AllVerbs_Return501(t *testing.T) {
cases := []struct {
name string
method string
invoke func(c *gin.Context)
}{
{
name: "ListFiles",
method: "GET",
invoke: func(c *gin.Context) { (&TemplatesHandler{}).ListFiles(c) },
},
{
name: "ReadFile",
method: "GET",
invoke: func(c *gin.Context) { (&TemplatesHandler{}).ReadFile(c) },
},
{
name: "WriteFile",
method: "PUT",
invoke: func(c *gin.Context) { (&TemplatesHandler{}).WriteFile(c) },
},
{
name: "DeleteFile",
method: "DELETE",
invoke: func(c *gin.Context) { (&TemplatesHandler{}).DeleteFile(c) },
},
}
for _, tc := range cases {
t.Run(tc.name, func(t *testing.T) {
w := httptest.NewRecorder()
c, _ := gin.CreateTestContext(w)
c.Params = gin.Params{
{Key: "id", Value: "ws-stub"},
// Path param without leading slash so DeleteFile's
// filepath.IsAbs guard doesn't 400 before the root
// dispatch runs. The List/Read/Write paths strip the
// leading slash themselves and accept either form.
{Key: "path", Value: "notes.md"},
}
// WriteFile binds JSON; provide a minimal valid body so the
// short-circuit isn't masked by the bind-error path.
var body string
if tc.method == "PUT" {
body = `{"content":"x"}`
}
c.Request = httptest.NewRequest(
tc.method,
"/workspaces/ws-stub/files/notes.md?root=/agent-home",
strings.NewReader(body),
)
if body != "" {
c.Request.Header.Set("Content-Type", "application/json")
}
tc.invoke(c)
if w.Code != http.StatusNotImplemented {
t.Fatalf("expected 501, got %d: %s", w.Code, w.Body.String())
}
if !strings.Contains(w.Body.String(), "/agent-home not implemented") {
t.Errorf("body should contain canonical stub message; got %s", w.Body.String())
}
if strings.Contains(w.Body.String(), "workspace not found") {
t.Errorf("stub leaked through to DB lookup; body=%s", w.Body.String())
}
})
}
}
@@ -18,11 +18,35 @@ import (
)
// allowedRoots are the container paths that the Files API can browse.
//
// `/agent-home` (added 2026-05-15, internal#425 RFC) is the container's
// own $HOME — `/root` for openclaw, `/home/agent` for claude-code/hermes
// — browsed via `docker exec` rather than host-side `find`. The
// dispatch is stubbed today (returns 501); full implementation lands in
// Phase 2b of the RFC. The allowedRoots key is added now so the canvas
// can design its root-selector UI against the final shape and the
// stub-vs-full transition is server-side only.
var allowedRoots = map[string]bool{
"/configs": true,
"/workspace": true,
"/home": true,
"/plugins": true,
"/configs": true,
"/workspace": true,
"/home": true,
"/plugins": true,
"/agent-home": true,
}
// agentHomeStubMessage is the body returned by every Files API verb
// when `?root=/agent-home` is requested before Phase 2b lands. Keep the
// status code 501 (Not Implemented) — the route exists, the verb is
// understood, but the handler is unimplemented. Distinguishes from
// 400/404 so a canvas behind a less-current server can render a clean
// "feature pending" state instead of a generic error.
const agentHomeStubMessage = "/agent-home not implemented yet (internal#425 RFC Phase 2b — docker-exec backend pending)"
// isAgentHomeStubRequest returns true when the request targets the
// stubbed /agent-home root. Centralised so every verb in this file
// short-circuits with the same response shape.
func isAgentHomeStubRequest(rootPath string) bool {
return rootPath == "/agent-home"
}
// maxUploadFiles limits the number of files in a single import/replace.
@@ -219,7 +243,14 @@ func (h *TemplatesHandler) ListFiles(c *gin.Context) {
// ?depth= — max depth to recurse (default: 1, max: 5)
rootPath := c.DefaultQuery("root", "/configs")
if !allowedRoots[rootPath] {
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins"})
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins, /agent-home"})
return
}
// /agent-home dispatch is stubbed pre-Phase-2b. Short-circuit before
// the DB lookup + EIC dance so a canvas exercising the new root key
// gets a clean 501 instead of a half-effort response.
if isAgentHomeStubRequest(rootPath) {
c.JSON(http.StatusNotImplemented, gin.H{"error": agentHomeStubMessage})
return
}
subPath := c.DefaultQuery("path", "")
@@ -383,7 +414,11 @@ func (h *TemplatesHandler) ReadFile(c *gin.Context) {
ctx := c.Request.Context()
rootPath := c.DefaultQuery("root", "/configs")
if !allowedRoots[rootPath] {
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins"})
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins, /agent-home"})
return
}
if isAgentHomeStubRequest(rootPath) {
c.JSON(http.StatusNotImplemented, gin.H{"error": agentHomeStubMessage})
return
}
@@ -496,7 +531,11 @@ func (h *TemplatesHandler) WriteFile(c *gin.Context) {
ctx := c.Request.Context()
rootPath := c.DefaultQuery("root", "/configs")
if !allowedRoots[rootPath] {
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins"})
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins, /agent-home"})
return
}
if isAgentHomeStubRequest(rootPath) {
c.JSON(http.StatusNotImplemented, gin.H{"error": agentHomeStubMessage})
return
}
var wsName, instanceID, runtime string
@@ -573,7 +612,11 @@ func (h *TemplatesHandler) DeleteFile(c *gin.Context) {
ctx := c.Request.Context()
rootPath := c.DefaultQuery("root", "/configs")
if !allowedRoots[rootPath] {
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins"})
c.JSON(http.StatusBadRequest, gin.H{"error": "root must be one of: /configs, /workspace, /home, /plugins, /agent-home"})
return
}
if isAgentHomeStubRequest(rootPath) {
c.JSON(http.StatusNotImplemented, gin.H{"error": agentHomeStubMessage})
return
}
var wsName, instanceID, runtime string
@@ -10,8 +10,20 @@ import (
"github.com/Molecule-AI/molecule-monorepo/platform/internal/db"
"github.com/Molecule-AI/molecule-monorepo/platform/internal/wsauth"
"github.com/gin-gonic/gin"
"github.com/google/uuid"
)
// validWorkspaceID returns true when id is a syntactically valid UUID.
// workspace_id is a `uuid` column; passing a non-UUID (e.g. the canvas
// "global" sentinel sent when no node is selected) makes Postgres raise
// `invalid input syntax for type uuid`, which previously leaked as an
// opaque 500. Reject up front with a clean 400 instead. Mirrors the
// uuid.Parse guard already used in handlers/activity.go.
func validWorkspaceID(id string) bool {
_, err := uuid.Parse(id)
return err == nil
}
// TokenHandler exposes user-facing token management for workspaces.
// Routes: GET/POST/DELETE /workspaces/:id/tokens (behind WorkspaceAuth).
type TokenHandler struct{}
@@ -31,6 +43,10 @@ type tokenListItem struct {
// never the plaintext or hash).
func (h *TokenHandler) List(c *gin.Context) {
workspaceID := c.Param("id")
if !validWorkspaceID(workspaceID) {
c.JSON(http.StatusBadRequest, gin.H{"error": "invalid workspace id"})
return
}
limit := 50
if v := c.Query("limit"); v != "" {
@@ -53,6 +69,7 @@ func (h *TokenHandler) List(c *gin.Context) {
LIMIT $2 OFFSET $3
`, workspaceID, limit, offset)
if err != nil {
log.Printf("tokens: list query failed for workspace %s: %v", workspaceID, err)
c.JSON(http.StatusInternalServerError, gin.H{"error": "failed to list tokens"})
return
}
@@ -85,6 +102,10 @@ const maxTokensPerWorkspace = 50
// exactly once in the response — it cannot be recovered afterwards.
func (h *TokenHandler) Create(c *gin.Context) {
workspaceID := c.Param("id")
if !validWorkspaceID(workspaceID) {
c.JSON(http.StatusBadRequest, gin.H{"error": "invalid workspace id"})
return
}
// Rate limit: max active tokens per workspace
var count int
@@ -117,6 +138,10 @@ func (h *TokenHandler) Create(c *gin.Context) {
func (h *TokenHandler) Revoke(c *gin.Context) {
workspaceID := c.Param("id")
tokenID := c.Param("tokenId")
if !validWorkspaceID(workspaceID) {
c.JSON(http.StatusBadRequest, gin.H{"error": "invalid workspace id"})
return
}
result, err := db.DB.ExecContext(c.Request.Context(), `
UPDATE workspace_auth_tokens
@@ -41,6 +41,15 @@ import (
func init() { gin.SetMode(gin.TestMode) }
// Workspace IDs are validated as UUIDs up front (tokens.go validWorkspaceID),
// so handler tests must pass syntactically valid UUIDs. Fixed values keep
// sqlmock WithArgs assertions deterministic.
const (
wsUUID1 = "11111111-1111-1111-1111-111111111111"
wsUUID2 = "22222222-2222-2222-2222-222222222222"
wsUUID3 = "33333333-3333-3333-3333-333333333333"
)
// withMockDB swaps `db.DB` for a sqlmock and returns the mock plus a
// restore func. Tests use this in place of setupTokenTestDB which
// skips on a missing real DB.
@@ -81,13 +90,13 @@ func TestTokenHandler_List_HappyPath(t *testing.T) {
created := time.Date(2026, 4, 1, 12, 0, 0, 0, time.UTC)
last := created.Add(time.Hour)
mock.ExpectQuery(`SELECT id, prefix, created_at, last_used_at\s+FROM workspace_auth_tokens`).
WithArgs("ws-1", 50, 0).
WithArgs(wsUUID1, 50, 0).
WillReturnRows(sqlmock.NewRows([]string{"id", "prefix", "created_at", "last_used_at"}).
AddRow("tok-1", "abc12345", created, last).
AddRow("tok-2", "def67890", created, nil))
w := makeReq(t, NewTokenHandler().List, "GET",
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: "ws-1"}})
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: wsUUID1}})
if w.Code != http.StatusOK {
t.Fatalf("expected 200, got %d: %s", w.Code, w.Body.String())
@@ -121,7 +130,7 @@ func TestTokenHandler_List_EmptyResult(t *testing.T) {
WillReturnRows(sqlmock.NewRows([]string{"id", "prefix", "created_at", "last_used_at"}))
w := makeReq(t, NewTokenHandler().List, "GET",
"/workspaces/ws-2/tokens", gin.Params{{Key: "id", Value: "ws-2"}})
"/workspaces/ws-2/tokens", gin.Params{{Key: "id", Value: wsUUID2}})
if w.Code != http.StatusOK {
t.Fatalf("expected 200 on empty list, got %d", w.Code)
@@ -146,7 +155,7 @@ func TestTokenHandler_List_QueryError(t *testing.T) {
WillReturnError(errors.New("connection refused"))
w := makeReq(t, NewTokenHandler().List, "GET",
"/workspaces/ws-3/tokens", gin.Params{{Key: "id", Value: "ws-3"}})
"/workspaces/ws-3/tokens", gin.Params{{Key: "id", Value: wsUUID3}})
if w.Code != http.StatusInternalServerError {
t.Errorf("query error must surface as 500, got %d", w.Code)
@@ -158,13 +167,13 @@ func TestTokenHandler_List_RespectsLimit(t *testing.T) {
defer cleanup()
mock.ExpectQuery(`SELECT id, prefix, created_at, last_used_at`).
WithArgs("ws-1", 10, 5).
WithArgs(wsUUID1, 10, 5).
WillReturnRows(sqlmock.NewRows([]string{"id", "prefix", "created_at", "last_used_at"}))
w := httptest.NewRecorder()
c, _ := gin.CreateTestContext(w)
c.Request = httptest.NewRequest("GET", "/workspaces/ws-1/tokens?limit=10&offset=5", nil)
c.Params = gin.Params{{Key: "id", Value: "ws-1"}}
c.Params = gin.Params{{Key: "id", Value: wsUUID1}}
NewTokenHandler().List(c)
if w.Code != http.StatusOK {
@@ -186,7 +195,7 @@ func TestTokenHandler_List_ScanError(t *testing.T) {
AddRow("tok-1", "abc", "not-a-timestamp", nil))
w := makeReq(t, NewTokenHandler().List, "GET",
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: "ws-1"}})
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: wsUUID1}})
if w.Code != http.StatusInternalServerError {
t.Errorf("scan error must surface as 500, got %d: %s", w.Code, w.Body.String())
@@ -201,11 +210,11 @@ func TestTokenHandler_Create_RateLimited(t *testing.T) {
// Count query returns 50 (== max) → 429.
mock.ExpectQuery(`SELECT COUNT\(\*\) FROM workspace_auth_tokens`).
WithArgs("ws-1").
WithArgs(wsUUID1).
WillReturnRows(sqlmock.NewRows([]string{"count"}).AddRow(50))
w := makeReq(t, NewTokenHandler().Create, "POST",
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: "ws-1"}})
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: wsUUID1}})
if w.Code != http.StatusTooManyRequests {
t.Errorf("max active tokens should 429, got %d", w.Code)
@@ -225,7 +234,7 @@ func TestTokenHandler_Create_IssueFails(t *testing.T) {
WillReturnError(errors.New("disk full"))
w := makeReq(t, NewTokenHandler().Create, "POST",
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: "ws-1"}})
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: wsUUID1}})
if w.Code != http.StatusInternalServerError {
t.Errorf("IssueToken DB error must 500, got %d", w.Code)
@@ -242,7 +251,7 @@ func TestTokenHandler_Create_HappyPath(t *testing.T) {
WillReturnResult(sqlmock.NewResult(1, 1))
w := makeReq(t, NewTokenHandler().Create, "POST",
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: "ws-1"}})
"/workspaces/ws-1/tokens", gin.Params{{Key: "id", Value: wsUUID1}})
if w.Code != http.StatusCreated {
t.Fatalf("expected 201, got %d: %s", w.Code, w.Body.String())
@@ -257,7 +266,7 @@ func TestTokenHandler_Create_HappyPath(t *testing.T) {
if body.AuthToken == "" {
t.Errorf("auth_token must be present and non-empty in response")
}
if body.WorkspaceID != "ws-1" {
if body.WorkspaceID != wsUUID1 {
t.Errorf("workspace_id mismatch: %q", body.WorkspaceID)
}
}
@@ -269,12 +278,12 @@ func TestTokenHandler_Revoke_HappyPath(t *testing.T) {
defer cleanup()
mock.ExpectExec(`UPDATE workspace_auth_tokens\s+SET revoked_at = now\(\)`).
WithArgs("tok-1", "ws-1").
WithArgs("tok-1", wsUUID1).
WillReturnResult(sqlmock.NewResult(0, 1))
w := makeReq(t, NewTokenHandler().Revoke, "DELETE",
"/workspaces/ws-1/tokens/tok-1", gin.Params{
{Key: "id", Value: "ws-1"},
{Key: "id", Value: wsUUID1},
{Key: "tokenId", Value: "tok-1"},
})
@@ -289,12 +298,12 @@ func TestTokenHandler_Revoke_NotFound(t *testing.T) {
// 0 rows affected → token not found OR already revoked.
mock.ExpectExec(`UPDATE workspace_auth_tokens`).
WithArgs("tok-ghost", "ws-1").
WithArgs("tok-ghost", wsUUID1).
WillReturnResult(sqlmock.NewResult(0, 0))
w := makeReq(t, NewTokenHandler().Revoke, "DELETE",
"/workspaces/ws-1/tokens/tok-ghost", gin.Params{
{Key: "id", Value: "ws-1"},
{Key: "id", Value: wsUUID1},
{Key: "tokenId", Value: "tok-ghost"},
})
@@ -312,7 +321,7 @@ func TestTokenHandler_Revoke_DBError(t *testing.T) {
w := makeReq(t, NewTokenHandler().Revoke, "DELETE",
"/workspaces/ws-1/tokens/tok-1", gin.Params{
{Key: "id", Value: "ws-1"},
{Key: "id", Value: wsUUID1},
{Key: "tokenId", Value: "tok-1"},
})
@@ -321,6 +330,59 @@ func TestTokenHandler_Revoke_DBError(t *testing.T) {
}
}
// ---- UUID validation (regression: "global" sentinel 500) ------------
// The canvas Settings → Workspace Tokens tab sent the literal sentinel
// "global" as the workspace id when no node was selected. workspace_id
// is a `uuid` column, so the query raised
// `invalid input syntax for type uuid: "global"` which leaked as an
// opaque 500. List/Create/Revoke now reject any non-UUID id with a
// clean 400 before touching the DB. No DB expectation is set on the
// mock — a DB hit would fail ExpectationsWereMet, proving short-circuit.
func TestTokenHandler_RejectsNonUUIDWorkspaceID(t *testing.T) {
h := NewTokenHandler()
cases := []struct {
name string
run func(c *gin.Context)
method string
params gin.Params
}{
{"List", h.List, "GET", gin.Params{{Key: "id", Value: "global"}}},
{"Create", h.Create, "POST", gin.Params{{Key: "id", Value: "global"}}},
{"Revoke", h.Revoke, "DELETE", gin.Params{
{Key: "id", Value: "global"},
{Key: "tokenId", Value: "tok-1"},
}},
}
for _, tc := range cases {
t.Run(tc.name, func(t *testing.T) {
mock, cleanup := withMockDB(t)
defer cleanup()
w := makeReq(t, tc.run, tc.method,
"/workspaces/global/tokens", tc.params)
if w.Code != http.StatusBadRequest {
t.Fatalf("%s with non-UUID id must 400, got %d: %s",
tc.name, w.Code, w.Body.String())
}
var body struct {
Error string `json:"error"`
}
_ = json.Unmarshal(w.Body.Bytes(), &body)
if body.Error != "invalid workspace id" {
t.Errorf("%s: want error=%q, got %q",
tc.name, "invalid workspace id", body.Error)
}
// No query/exec was expected → if the handler hit the DB
// this fails, proving the guard short-circuits before SQL.
if err := mock.ExpectationsWereMet(); err != nil {
t.Errorf("%s leaked a DB call past the uuid guard: %v", tc.name, err)
}
})
}
}
// Compile-time noise removal: the imports list pulls in the sql /
// driver packages and the silenced ctx so a future scenario that
// needs them doesn't have to re-add the import. Documented here so
@@ -11,6 +11,7 @@ import (
"github.com/Molecule-AI/molecule-monorepo/platform/internal/db"
"github.com/Molecule-AI/molecule-monorepo/platform/internal/wsauth"
"github.com/gin-gonic/gin"
"github.com/google/uuid"
)
func init() { gin.SetMode(gin.TestMode) }
@@ -167,11 +168,14 @@ func TestTokenHandler_RevokeWrongWorkspace(t *testing.T) {
h := NewTokenHandler()
// Try to revoke with a different workspace ID — should 404
// Try to revoke with a different (valid-UUID) workspace ID that does
// not own the token — should 404. A valid UUID is required so this
// exercises the ownership branch, not the up-front uuid-shape 400.
otherWS := uuid.NewString()
w := httptest.NewRecorder()
c, _ := gin.CreateTestContext(w)
c.Params = gin.Params{{Key: "id", Value: "wrong-workspace-id"}, {Key: "tokenId", Value: tokenID}}
c.Request = httptest.NewRequest("DELETE", "/workspaces/wrong/tokens/"+tokenID, nil)
c.Params = gin.Params{{Key: "id", Value: otherWS}, {Key: "tokenId", Value: tokenID}}
c.Request = httptest.NewRequest("DELETE", "/workspaces/"+otherWS+"/tokens/"+tokenID, nil)
h.Revoke(c)
if w.Code != http.StatusNotFound {
@@ -80,6 +80,15 @@ type WorkspaceHandler struct {
asyncWG sync.WaitGroup
}
// newHandlerHook, when non-nil, is invoked for every WorkspaceHandler
// created via NewWorkspaceHandler. It is nil in production (zero cost);
// the test harness sets it so setupTestDB can drain every handler's
// in-flight async goroutines before swapping the global db.DB. Without
// this, a detached restart goroutine (maybeMarkContainerDead ->
// goAsync(RestartByID) -> runRestartCycle reads db.DB) races the
// db.DB restore in another test's t.Cleanup.
var newHandlerHook func(*WorkspaceHandler)
func (h *WorkspaceHandler) goAsync(fn func()) {
h.asyncWG.Add(1)
go func() {
@@ -108,6 +117,9 @@ func NewWorkspaceHandler(b events.EventEmitter, p *provisioner.Provisioner, plat
if p != nil {
h.provisioner = p
}
if newHandlerHook != nil {
newHandlerHook(h)
}
return h
}
@@ -237,10 +237,10 @@ func (h *WorkspaceHandler) Restart(c *gin.Context) {
// the silent-drop bugs PRs #2811/#2824 closed). RestartWorkspaceAuto
// enforces CP-FIRST ordering matching the other dispatchers — see
// docs/architecture/backends.md.
go func() {
h.goAsync(func() {
h.RestartWorkspaceAutoOpts(context.Background(), id, templatePath, configFiles, payload, resetClaudeSession)
}()
go h.sendRestartContext(id, restartData)
})
h.goAsync(func() { h.sendRestartContext(id, restartData) })
c.JSON(http.StatusOK, gin.H{"status": "provisioning", "config_dir": configLabel, "reset_session": resetClaudeSession})
}
@@ -610,7 +610,9 @@ func (h *WorkspaceHandler) runRestartCycle(workspaceID string) {
h.provisionWorkspaceAutoSync(workspaceID, "", nil, payload)
// sendRestartContext is a one-way notification to the new container; safe
// to fire async — the next restart cycle won't depend on it completing.
go h.sendRestartContext(workspaceID, restartData)
// Tracked via goAsync so the test harness can drain it before the
// global db.DB swap (sendRestartContext reads db.DB).
h.goAsync(func() { h.sendRestartContext(workspaceID, restartData) })
}
// Pause handles POST /workspaces/:id/pause
@@ -178,12 +178,21 @@ func (p *CPProvisioner) Start(ctx context.Context, cfg WorkspaceConfig) (string,
// /admin/liveness and other admin-gated platform endpoints (core#831).
// p.adminToken is read from os.Getenv("ADMIN_TOKEN") at provisioner creation;
// it is also used for CP→platform HTTP auth but those are separate concerns.
env := cfg.EnvVars
if p.adminToken != "" {
env = make(map[string]string, len(cfg.EnvVars)+1)
for k, v := range cfg.EnvVars {
env[k] = v
//
// Forensic #145 hardening: tenant workspaces run on EC2 via this path, so
// the SCM-write-token denylist (see buildContainerEnv) is enforced here
// too. Always build a filtered copy — never pass cfg.EnvVars through
// verbatim — so a latent persona-merged GITEA_TOKEN can't reach the
// tenant container regardless of whether ADMIN_TOKEN is set.
env := make(map[string]string, len(cfg.EnvVars)+1)
for k, v := range cfg.EnvVars {
if isSCMWriteTokenKey(k) {
log.Printf("CPProvisioner.Start: dropped SCM-write credential %q from tenant workspace env (forensic #145 guard)", k)
continue
}
env[k] = v
}
if p.adminToken != "" {
env["ADMIN_TOKEN"] = p.adminToken
}
// Collect template files and generated configs, with OFFSEC-010 guards:
@@ -343,6 +352,7 @@ func collectCPConfigFiles(cfg WorkspaceConfig) (map[string]string, error) {
}
return files, nil
}
// Stop terminates the workspace's EC2 instance via the control plane.
//
// Looks up the actual EC2 instance_id from the workspaces table before
@@ -497,7 +507,9 @@ func (p *CPProvisioner) IsRunning(ctx context.Context, workspaceID string) (bool
// Don't leak the body — upstream errors may echo headers.
return true, fmt.Errorf("cp provisioner: status: unexpected %d", resp.StatusCode)
}
var result struct{ State string `json:"state"` }
var result struct {
State string `json:"state"`
}
// Cap body read at 64 KiB for parity with Start — a misconfigured
// or compromised CP streaming a huge body could otherwise exhaust
// memory in this hot path (called reactively per-request from
@@ -591,6 +591,28 @@ func ValidateWorkspaceAccess(access, workspacePath string) error {
}
}
// scmWriteTokenKeys is the explicit denylist of environment variable names
// that carry a Git SCM *write* credential (push / merge / approve). These
// must never reach a tenant workspace container — see the forensic #145
// rationale in buildContainerEnv. Kept as an exact-match set rather than a
// substring/prefix heuristic so the guard is auditable and can't silently
// over-strip a legitimately-named var.
var scmWriteTokenKeys = map[string]struct{}{
"GITEA_TOKEN": {},
"GITHUB_TOKEN": {},
"GH_TOKEN": {}, // gh CLI honours GH_TOKEN as a GITHUB_TOKEN alias
"GITLAB_TOKEN": {},
"GL_TOKEN": {}, // glab CLI alias
"BITBUCKET_TOKEN": {},
}
// isSCMWriteTokenKey reports whether an env var name is a known Git SCM
// write credential that must be stripped from tenant workspace env.
func isSCMWriteTokenKey(key string) bool {
_, ok := scmWriteTokenKeys[key]
return ok
}
// buildContainerEnv assembles the initial environment variables injected
// into every workspace container.
//
@@ -627,6 +649,21 @@ func buildContainerEnv(cfg WorkspaceConfig) []string {
env = append(env, fmt.Sprintf("AWARENESS_URL=%s", cfg.AwarenessURL))
}
for k, v := range cfg.EnvVars {
// Forensic #145 hardening: tenant workspace containers run
// agent-controlled code and must NEVER receive a Git SCM *write*
// credential. Without merge/approve creds in-container the
// two-eyes review gate is structurally self-bypass-proof — an
// agent that forges an approval has no token to act on it. A
// latent path exists (loadPersonaEnvFile merges a per-role
// persona `GITEA_TOKEN` into cfg.EnvVars when MOLECULE_PERSONA_ROOT
// is set on a tenant host); it is inert today (persona dirs are
// operator-host-only) but unguarded. Strip SCM-write tokens here
// by construction so the invariant holds regardless of whether
// that path ever becomes reachable.
if isSCMWriteTokenKey(k) {
log.Printf("buildContainerEnv: dropped SCM-write credential %q from workspace env (forensic #145 guard)", k)
continue
}
env = append(env, fmt.Sprintf("%s=%s", k, v))
}
// Inject ADMIN_TOKEN from the platform server's environment so workspace
@@ -636,10 +636,15 @@ func TestBuildContainerEnv_AwarenessOnlyWhenBothSet(t *testing.T) {
}
func TestBuildContainerEnv_CustomEnvVarsAppended(t *testing.T) {
// NOTE: this test previously asserted GITHUB_TOKEN passed through
// verbatim. That assertion encoded the forensic #145 latent leak as
// expected behavior. Post-guard, ordinary custom env still flows but
// SCM-write credentials are stripped — see
// TestBuildContainerEnv_StripsSCMWriteTokens for the negative assertion.
cfg := WorkspaceConfig{
WorkspaceID: "ws-x",
PlatformURL: "http://localhost:8080",
EnvVars: map[string]string{"CUSTOM": "value", "GITHUB_TOKEN": "fake-token-for-test"},
EnvVars: map[string]string{"CUSTOM": "value", "ANTHROPIC_API_KEY": "sk-not-an-scm-token"},
}
env := buildContainerEnv(cfg)
seen := map[string]string{}
@@ -652,8 +657,8 @@ func TestBuildContainerEnv_CustomEnvVarsAppended(t *testing.T) {
if seen["CUSTOM"] != "value" {
t.Errorf("CUSTOM env missing, got env=%v", env)
}
if seen["GITHUB_TOKEN"] != "fake-token-for-test" {
t.Errorf("GITHUB_TOKEN env missing, got env=%v", env)
if seen["ANTHROPIC_API_KEY"] != "sk-not-an-scm-token" {
t.Errorf("non-SCM custom env must still pass through, got env=%v", env)
}
// Built-in defaults still present
if seen["MOLECULE_URL"] == "" {
@@ -661,6 +666,129 @@ func TestBuildContainerEnv_CustomEnvVarsAppended(t *testing.T) {
}
}
// ---------- forensic #145: SCM-write-token denylist guard ----------
// TestBuildContainerEnv_StripsSCMWriteTokens is the core negative
// assertion: a tenant workspace env constructed via buildContainerEnv MUST
// NOT contain any Git SCM *write* credential, regardless of how it got into
// cfg.EnvVars. This proves the two-eyes review gate stays structurally
// self-bypass-proof — an agent in-container has no merge/approve token to
// act on a forged approval. See forensic #145.
//
// This test FAILS on the pre-guard code (where buildContainerEnv passed
// cfg.EnvVars through verbatim) and PASSES once the denylist filter is in
// place — i.e. the guard is proven by construction, not by environment
// accident.
func TestBuildContainerEnv_StripsSCMWriteTokens(t *testing.T) {
scmTokens := []string{
"GITEA_TOKEN", "GITHUB_TOKEN", "GH_TOKEN",
"GITLAB_TOKEN", "GL_TOKEN", "BITBUCKET_TOKEN",
}
t.Run("normal path — SCM tokens explicitly set in EnvVars", func(t *testing.T) {
envVars := map[string]string{"CUSTOM": "ok", "ANTHROPIC_API_KEY": "sk-keep"}
for _, k := range scmTokens {
envVars[k] = "leaked-write-credential-" + k
}
cfg := WorkspaceConfig{
WorkspaceID: "ws-tenant",
PlatformURL: "http://localhost:8080",
Tier: 2,
EnvVars: envVars,
}
assertNoSCMWriteToken(t, buildContainerEnv(cfg), scmTokens)
// Sanity: non-SCM custom env is NOT collateral-damaged by the filter.
if !envContains(buildContainerEnv(cfg), "CUSTOM=ok") {
t.Errorf("filter must not strip non-SCM custom env")
}
if !envContains(buildContainerEnv(cfg), "ANTHROPIC_API_KEY=sk-keep") {
t.Errorf("filter must not strip non-SCM API keys")
}
})
t.Run("persona-file path — simulates loadPersonaEnvFile merge", func(t *testing.T) {
// The latent path: handlers.loadPersonaEnvFile() merges a per-role
// persona env file (carrying GITEA_USER, GITEA_TOKEN, …) into the
// workspace env map when MOLECULE_PERSONA_ROOT is set on a tenant
// host. We can't invoke that cross-package helper here, but its
// observable effect is exactly "a GITEA_TOKEN appears in
// cfg.EnvVars". Constructing that condition directly proves the
// guard holds even if the latent path becomes reachable.
cfg := WorkspaceConfig{
WorkspaceID: "ws-tenant",
PlatformURL: "http://localhost:8080",
Tier: 2,
EnvVars: map[string]string{
// Persona identity fields that are SAFE to keep (read-only
// identity, not a write credential):
"GITEA_USER": "backend-engineer",
"GITEA_USER_EMAIL": "backend-engineer@agents.moleculesai.app",
// The credential that must be stripped:
"GITEA_TOKEN": "persona-merged-write-pat",
"GITEA_TOKEN_SCOPES": "write:repository",
},
}
got := buildContainerEnv(cfg)
assertNoSCMWriteToken(t, got, scmTokens)
// Non-credential persona identity may still flow through — only the
// write token is the denied surface.
if !envContains(got, "GITEA_USER=backend-engineer") {
t.Errorf("non-credential persona identity (GITEA_USER) should not be stripped")
}
})
}
// TestCPProvisionerEnv_StripsSCMWriteTokens covers the tenant-EC2 path:
// CPProvisioner.Start builds the env map the control plane forwards to the
// EC2 workspace container. The same forensic #145 denylist must hold there.
func TestCPProvisionerEnv_StripsSCMWriteTokens(t *testing.T) {
// isSCMWriteTokenKey is the single source of truth shared by both
// buildContainerEnv (local Docker) and CPProvisioner.Start (tenant EC2).
// Assert it classifies every known SCM-write var as denied and leaves
// ordinary / read-only-identity vars alone.
for _, k := range []string{
"GITEA_TOKEN", "GITHUB_TOKEN", "GH_TOKEN",
"GITLAB_TOKEN", "GL_TOKEN", "BITBUCKET_TOKEN",
} {
if !isSCMWriteTokenKey(k) {
t.Errorf("isSCMWriteTokenKey(%q) = false, want true (SCM-write credential must be denied)", k)
}
}
for _, k := range []string{
"GITEA_USER", "GITEA_USER_EMAIL", "ANTHROPIC_API_KEY",
"CUSTOM", "PLATFORM_URL", "ADMIN_TOKEN", "",
} {
if isSCMWriteTokenKey(k) {
t.Errorf("isSCMWriteTokenKey(%q) = true, want false (must not over-strip non-SCM env)", k)
}
}
}
func assertNoSCMWriteToken(t *testing.T, env []string, scmTokens []string) {
t.Helper()
for _, e := range env {
key := e
if i := strings.IndexByte(e, '='); i >= 0 {
key = e[:i]
}
for _, banned := range scmTokens {
if key == banned {
t.Errorf("SCM-write credential %q leaked into workspace env (forensic #145 invariant violated): %q", banned, e)
}
}
}
}
func envContains(env []string, want string) bool {
for _, e := range env {
if e == want {
return true
}
}
return false
}
// ---------- buildWorkspaceMount — #65 workspace_access ----------
func TestBuildWorkspaceMount_SelectionMatrix(t *testing.T) {
@@ -0,0 +1,226 @@
// Package secrets provides the canonical SSOT for credential-shaped
// regex patterns used by:
//
// - the CI `Secret scan` workflow (.gitea/workflows/secret-scan.yml)
// - the runtime's bundled pre-commit hook
// (molecule-ai-workspace-runtime/molecule_runtime/scripts/pre-commit-checks.sh)
// - the upcoming Phase 2b docker-exec Files API backend, which has
// to refuse to surface files whose path OR content matches a
// credential shape (RFC internal#425, Hongming 2026-05-15)
//
// Before this package, the same regex set lived as duplicate bash
// arrays in two unrelated repos; adding a pattern required editing
// both, and pattern drift was caught only via secret-scan workflow
// failures on PRs that had unrelated changes (#2090-class incident
// vector). Centralising in Go makes the Files API the SSOT, with the
// YAML + bash arrays generated/asserted from this package so drift
// is detected at CI time, not at exfiltration time.
//
// This file is Phase 2a of the internal#425 RFC. Phase 2b will import
// `Patterns` from `template_files_docker_exec.go` to gate
// `listFilesViaDockerExec` / `readFileViaDockerExec` against
// secret-shaped paths AND content. Until 2b lands, the package has
// one consumer: this package's own unit tests, which pin the regex
// strings so a refactor that drops or weakens one is caught here.
package secrets
import (
"fmt"
"regexp"
"sync"
)
// Pattern is one named credential shape — a human label plus the
// compiled regex. The label appears in CI error output ("matched:
// github-pat") so an operator can identify the family without seeing
// the actual matched bytes (echoing the bytes widens the blast radius
// per the secret-scan workflow's recovery prose).
type Pattern struct {
// Name is a short kebab-case identifier (e.g. "github-pat",
// "anthropic-api-key"). Stable across versions — consumers may
// switch on it.
Name string
// Description is a one-line human-readable explanation of what
// the pattern matches. Used in CI error messages and the Files
// API "<denied: secret-shape>" placeholder rationale.
Description string
// regexSource is the regex literal in Go-RE2 syntax. Stored as a
// string so the slice declaration below stays readable; compiled
// once via sync.Once into a *regexp.Regexp.
regexSource string
}
// Patterns is the canonical credential-shape regex set.
//
// Adding a pattern here:
//
// 1. Add a new Pattern{} entry below with a kebab-case Name, a
// one-line Description, and the regex literal. Anchor on a
// low-false-positive prefix.
// 2. Add a positive + negative test case in patterns_test.go.
// 3. Mirror the regex string into:
// a. .gitea/workflows/secret-scan.yml SECRET_PATTERNS array
// b. molecule-ai-workspace-runtime/molecule_runtime/scripts/pre-commit-checks.sh
// (or wait for the codegen target that consumes this slice — TBD
// follow-up; tracked in the Phase 2a PR description.)
//
// The order is: alphabetical within each provider family, families
// grouped by ecosystem (GitHub family, AI-provider family, chat
// family, cloud family). Keep this stable so diffs are reviewable.
var Patterns = []Pattern{
// --- GitHub token family ---
{
Name: "github-pat-classic",
Description: "GitHub personal access token (classic)",
regexSource: `ghp_[A-Za-z0-9]{36,}`,
},
{
Name: "github-app-installation-token",
Description: "GitHub App installation token (#2090 vector)",
regexSource: `ghs_[A-Za-z0-9]{36,}`,
},
{
Name: "github-oauth-user-to-server",
Description: "GitHub OAuth user-to-server token",
regexSource: `gho_[A-Za-z0-9]{36,}`,
},
{
Name: "github-oauth-user",
Description: "GitHub OAuth user token",
regexSource: `ghu_[A-Za-z0-9]{36,}`,
},
{
Name: "github-oauth-refresh",
Description: "GitHub OAuth refresh token",
regexSource: `ghr_[A-Za-z0-9]{36,}`,
},
{
Name: "github-pat-fine-grained",
Description: "GitHub fine-grained personal access token",
regexSource: `github_pat_[A-Za-z0-9_]{82,}`,
},
// --- AI-provider API key family ---
{
Name: "anthropic-api-key",
Description: "Anthropic API key",
regexSource: `sk-ant-[A-Za-z0-9_-]{40,}`,
},
{
Name: "openai-project-key",
Description: "OpenAI project API key",
regexSource: `sk-proj-[A-Za-z0-9_-]{40,}`,
},
{
Name: "openai-service-account-key",
Description: "OpenAI service-account API key",
regexSource: `sk-svcacct-[A-Za-z0-9_-]{40,}`,
},
{
Name: "minimax-api-key",
Description: "MiniMax API key (F1088 vector)",
regexSource: `sk-cp-[A-Za-z0-9_-]{60,}`,
},
// --- Chat-platform token family ---
{
Name: "slack-token",
Description: "Slack token (xoxb/xoxa/xoxp/xoxr/xoxs)",
regexSource: `xox[baprs]-[A-Za-z0-9-]{20,}`,
},
// --- Cloud-provider credential family ---
{
Name: "aws-access-key-id",
Description: "AWS access key ID",
regexSource: `AKIA[0-9A-Z]{16}`,
},
{
Name: "aws-sts-temp-access-key-id",
Description: "AWS STS temporary access key ID",
regexSource: `ASIA[0-9A-Z]{16}`,
},
}
// compiledOnce protects the lazy build of compiledPatterns. We compile
// lazily so package init is cheap; callers pay only on first match
// (typically once per workspace-server boot).
var (
compiledOnce sync.Once
compiledPatterns []*compiledPattern
compileErr error
)
type compiledPattern struct {
Name string
Description string
Re *regexp.Regexp
}
// compileAll compiles every Pattern.regexSource into a *regexp.Regexp.
// Called once via compiledOnce. Any compile failure here is a build
// bug (the unit tests assert each regex compiles) — surfacing via
// returned error so callers don't panic in request handling.
func compileAll() {
out := make([]*compiledPattern, 0, len(Patterns))
for _, p := range Patterns {
re, err := regexp.Compile(p.regexSource)
if err != nil {
compileErr = fmt.Errorf("secrets: pattern %q failed to compile: %w", p.Name, err)
return
}
out = append(out, &compiledPattern{Name: p.Name, Description: p.Description, Re: re})
}
compiledPatterns = out
}
// ScanBytes returns a non-nil Match if any pattern matches anywhere
// inside b. Returns (nil, nil) on no match. Returns (nil, err) only
// if a regex in the package fails to compile — that's a build bug,
// not a runtime data issue.
//
// Match contains the pattern Name + Description so the caller can
// emit a path-or-content-denial rationale WITHOUT round-tripping the
// matched bytes (which would defeat the purpose). The matched bytes
// stay inside this function.
//
// The Files API Phase 2b backend will call ScanBytes on:
//
// - the absolute path string (catches a file literally named
// `ghs_abc.txt`)
// - the file content (catches a credential pasted into a workspace
// file by an agent or user — the Files API refuses to surface it
// and the canvas renders "<denied: secret-shape>")
//
// Ordering: patterns are tried in declaration order. First match
// wins. This means narrower patterns (e.g. `sk-svcacct-…`) should
// appear in `Patterns` before broader ones (`sk-…`) — today there's
// no overlap, so order is descriptive only.
func ScanBytes(b []byte) (*Match, error) {
compiledOnce.Do(compileAll)
if compileErr != nil {
return nil, compileErr
}
for _, cp := range compiledPatterns {
if cp.Re.Match(b) {
return &Match{Name: cp.Name, Description: cp.Description}, nil
}
}
return nil, nil
}
// ScanString is the string-input convenience wrapper around ScanBytes.
// Identical semantics — the body never copies, []byte(s) is a
// zero-copy reinterpret for the regex matcher.
func ScanString(s string) (*Match, error) {
return ScanBytes([]byte(s))
}
// Match describes which pattern caught a value. Deliberately does
// NOT include the matched substring — callers must not echo it.
type Match struct {
// Name is the pattern's kebab-case identifier (e.g. "github-pat-classic").
Name string
// Description is the human-readable line for UI / log surfaces.
Description string
}
@@ -0,0 +1,189 @@
package secrets
import (
"strings"
"testing"
)
// TestEveryPatternCompiles pins that every Pattern.regexSource is a
// valid Go-RE2 expression. Without this, a bad regex would silently
// disable ScanBytes for everything after it (the lazy compile would
// set compileErr and ScanBytes would return that error every call).
func TestEveryPatternCompiles(t *testing.T) {
for _, p := range Patterns {
if p.Name == "" {
t.Errorf("pattern with empty Name: regex=%q", p.regexSource)
}
if p.Description == "" {
t.Errorf("pattern %q has empty Description", p.Name)
}
}
// Force compile + check error.
if _, err := ScanBytes([]byte("placeholder")); err != nil {
t.Fatalf("ScanBytes init failed: %v", err)
}
}
// TestNoDuplicateNames — a duplicate pattern Name would make the
// "first match wins" semantics surprising to readers and any caller
// switching on Match.Name (none today but adding the guard is cheap).
func TestNoDuplicateNames(t *testing.T) {
seen := map[string]bool{}
for _, p := range Patterns {
if seen[p.Name] {
t.Errorf("duplicate pattern Name: %q", p.Name)
}
seen[p.Name] = true
}
}
// TestKnownPatternsAllPresent — pins which specific Name values are
// expected. A future refactor that renames or removes one without
// updating consumers (CI workflow, runtime pre-commit hook, Files
// API Phase 2b backend) would silently widen the leak surface.
// Failing here forces the rename to be intentional.
func TestKnownPatternsAllPresent(t *testing.T) {
expected := []string{
"github-pat-classic",
"github-app-installation-token",
"github-oauth-user-to-server",
"github-oauth-user",
"github-oauth-refresh",
"github-pat-fine-grained",
"anthropic-api-key",
"openai-project-key",
"openai-service-account-key",
"minimax-api-key",
"slack-token",
"aws-access-key-id",
"aws-sts-temp-access-key-id",
}
got := map[string]bool{}
for _, p := range Patterns {
got[p.Name] = true
}
for _, want := range expected {
if !got[want] {
t.Errorf("expected pattern %q missing from Patterns slice", want)
}
}
}
// TestPositiveMatches — for each pattern, supply a representative
// shape and assert ScanBytes returns a Match with the right Name.
// These are TEST FIXTURES, not real credentials — each is the
// pattern's prefix + a long-enough trailing run of placeholder chars.
// `EXAMPLE` is sprinkled in to make grep-finds in CI logs obviously
// fake to a human reader (matches saved memory
// feedback_assert_exact_not_substring: tighten by Name not body).
func TestPositiveMatches(t *testing.T) {
cases := []struct {
fixture string
expectedName string
}{
{"ghp_EXAMPLE111122223333444455556666777788889999", "github-pat-classic"},
{"ghs_EXAMPLE111122223333444455556666777788889999", "github-app-installation-token"},
{"gho_EXAMPLE111122223333444455556666777788889999", "github-oauth-user-to-server"},
{"ghu_EXAMPLE111122223333444455556666777788889999", "github-oauth-user"},
{"ghr_EXAMPLE111122223333444455556666777788889999", "github-oauth-refresh"},
{"github_pat_EXAMPLE" + strings.Repeat("1", 80), "github-pat-fine-grained"},
{"sk-ant-EXAMPLE" + strings.Repeat("1", 40), "anthropic-api-key"},
{"sk-proj-EXAMPLE" + strings.Repeat("1", 40), "openai-project-key"},
{"sk-svcacct-EXAMPLE" + strings.Repeat("1", 40), "openai-service-account-key"},
{"sk-cp-EXAMPLE" + strings.Repeat("1", 60), "minimax-api-key"},
{"xoxb-" + strings.Repeat("a", 25), "slack-token"},
{"xoxa-" + strings.Repeat("a", 25), "slack-token"},
// AWS regex requires [0-9A-Z]{16} — uppercase + digits only.
{"AKIA1234567890ABCDEF", "aws-access-key-id"},
{"ASIA1234567890ABCDEF", "aws-sts-temp-access-key-id"},
}
for _, tc := range cases {
t.Run(tc.expectedName, func(t *testing.T) {
m, err := ScanBytes([]byte(tc.fixture))
if err != nil {
t.Fatalf("ScanBytes(%q) errored: %v", tc.fixture, err)
}
if m == nil {
t.Fatalf("ScanBytes(%q) returned no match — expected %q", tc.fixture, tc.expectedName)
}
if m.Name != tc.expectedName {
t.Errorf("ScanBytes(%q) matched %q; expected %q", tc.fixture, m.Name, tc.expectedName)
}
})
}
}
// TestNegativeShapes — strings that look credential-adjacent but
// shouldn't match (too short, wrong prefix, missing trailing bytes).
// Failing here means a pattern is too loose, which would generate
// false-positive denial in Files API and false-positive workflow
// failures in CI.
func TestNegativeShapes(t *testing.T) {
cases := []string{
// Too-short variants — anchored on the length suffix.
"ghp_tooshort",
"ghs_alsoshort1234",
"github_pat_short",
"sk-ant-short",
"sk-cp-not-enough-bytes-here",
// Looks like one of the prefixes but isn't (different letter).
"gha_EXAMPLE_thirty_six_or_more_chars_here_xxx",
// Slack family — wrong letter after xox.
"xoxz-aaaaaaaaaaaaaaaaaaaaaaaaa",
// AWS-shaped but wrong length suffix.
"AKIATOOSHORT",
// Empty / whitespace.
"",
" ",
// Plain prose mentioning the prefix as part of a longer word.
"see also `ghp_HOWTO.md` in the repo",
}
for _, c := range cases {
t.Run(c, func(t *testing.T) {
m, err := ScanBytes([]byte(c))
if err != nil {
t.Fatalf("ScanBytes(%q) errored: %v", c, err)
}
if m != nil {
t.Errorf("ScanBytes(%q) unexpectedly matched %q", c, m.Name)
}
})
}
}
// TestScanString_NoOp — sanity-check ScanString is the zero-copy
// wrapper around ScanBytes. Without this, a future refactor that
// makes ScanString do its own thing (e.g. accidentally normalise
// case) would diverge silently.
func TestScanString_NoOp(t *testing.T) {
in := "ghp_EXAMPLE111122223333444455556666777788889999"
m1, err1 := ScanBytes([]byte(in))
if err1 != nil {
t.Fatalf("ScanBytes errored: %v", err1)
}
m2, err2 := ScanString(in)
if err2 != nil {
t.Fatalf("ScanString errored: %v", err2)
}
if m1 == nil || m2 == nil {
t.Fatalf("expected matches; got bytes=%+v string=%+v", m1, m2)
}
if m1.Name != m2.Name {
t.Errorf("ScanString and ScanBytes returned different Names: %q vs %q", m1.Name, m2.Name)
}
}
// TestMatch_NoRoundtrip — assert the Match struct does NOT include
// the matched substring as a field. Adding such a field would
// regress the "matched bytes never leave ScanBytes" invariant that
// makes this package safe to call from log/UI surfaces. This is a
// reflection-light contract test — checks the field names statically.
func TestMatch_NoRoundtrip(t *testing.T) {
var m Match
// If someone adds a `Matched string` (or similar) field, this
// test reads as the canonical place to update + reconsider.
_ = m.Name
_ = m.Description
// The two-field shape is part of the public contract; new fields
// require deliberation about whether they leak the secret value.
}
+6
View File
@@ -35,12 +35,14 @@ from a2a_tools import (
tool_commit_memory,
tool_delegate_task,
tool_delegate_task_async,
tool_get_runtime_identity,
tool_get_workspace_info,
tool_inbox_peek,
tool_inbox_pop,
tool_list_peers,
tool_recall_memory,
tool_send_message_to_user,
tool_update_agent_card,
tool_wait_for_message,
)
from platform_tools.registry import TOOLS as _PLATFORM_TOOL_SPECS
@@ -130,6 +132,10 @@ async def handle_tool_call(name: str, arguments: dict) -> str:
return await tool_get_workspace_info(
source_workspace_id=arguments.get("source_workspace_id") or None,
)
elif name == "get_runtime_identity":
return await tool_get_runtime_identity()
elif name == "update_agent_card":
return await tool_update_agent_card(arguments.get("card"))
elif name == "commit_memory":
return await tool_commit_memory(
arguments.get("content", ""),
+12
View File
@@ -167,3 +167,15 @@ from a2a_tools_inbox import ( # noqa: E402 (import after the top-of-module imp
tool_inbox_pop,
tool_wait_for_message,
)
# Identity tool handlers — extracted to a2a_tools_identity. Ports the
# two T4-tier MCP tools (``tool_get_runtime_identity`` +
# ``tool_update_agent_card``) from molecule-ai-workspace-runtime PR#17.
# That repo is mirror-only (reference_runtime_repo_is_mirror_only);
# this is the canonical edit point, and the wheel mirror is
# regenerated by publish-runtime.yml on merge.
from a2a_tools_identity import ( # noqa: E402 (import after the top-of-module imports)
tool_get_runtime_identity,
tool_update_agent_card,
)
+187
View File
@@ -0,0 +1,187 @@
"""Identity tool handlers — single-concern slice of the a2a_tools surface.
Owns the two MCP tools that close the T4-tier workspace owner-permission
gaps reported via the canvas:
* ``tool_get_runtime_identity`` — env-only; returns model, model_provider,
molecule_model, anthropic_base_url, tier, workspace_id, runtime
(ADAPTER_MODULE). No HTTP call. Always permitted by RBAC — even
read-only agents may know what model they are.
* ``tool_update_agent_card`` — POSTs the card to ``/registry/update-card``
with the workspace's own bearer (same auth path as ``tool_commit_memory``
via ``a2a_tools_rbac.auth_headers_for_heartbeat``). The platform
replaces the stored card and broadcasts an ``agent_card_updated``
event so the canvas reflects the new card live. Gated on
``memory.write`` capability via the existing RBAC permission map so
read-only roles can't silently rewrite the platform card.
Both originated as a port of molecule-ai-workspace-runtime PR#17
(``feat(mcp): add update_agent_card + get_runtime_identity tools``).
The mirror-only PR#17 was closed without merge per
``reference_runtime_repo_is_mirror_only``; the canonical edit point is
this monorepo at ``workspace/`` and the wheel mirror is regenerated
automatically by the publish-runtime workflow.
Imports the auth-header primitive from ``a2a_tools_rbac`` (iter 4a) —
NOT from ``a2a_tools`` — to avoid a circular import with the
kitchen-sink re-export module.
"""
from __future__ import annotations
import json
import os
from typing import Any
import httpx
from a2a_client import PLATFORM_URL
from a2a_tools_rbac import (
auth_headers_for_heartbeat as _auth_headers_for_heartbeat,
check_memory_write_permission as _check_memory_write_permission,
)
def _runtime_identity_payload() -> dict[str, Any]:
"""Build the identity dict — env-only, no I/O.
Factored out from ``tool_get_runtime_identity`` so tests can assert
against the exact key set without re-parsing JSON. The MCP tool
handler ``tool_get_runtime_identity`` is the only public caller in
production; tests call this helper directly.
"""
return {
"model": os.environ.get("MODEL", ""),
"model_provider": os.environ.get("MODEL_PROVIDER", ""),
"molecule_model": os.environ.get("MOLECULE_MODEL", ""),
"anthropic_base_url": os.environ.get("ANTHROPIC_BASE_URL", ""),
"tier": os.environ.get("TIER", ""),
"workspace_id": os.environ.get("WORKSPACE_ID", ""),
# Adapter module is the closest thing the runtime has to a
# "template slug" — e.g. "adapter" for claude-code-default,
# "hermes" for hermes-template, etc. Picked from
# $ADAPTER_MODULE env baked by each template's Dockerfile.
"runtime": os.environ.get("ADAPTER_MODULE", ""),
}
async def tool_get_runtime_identity() -> str:
"""Return this runtime's identity — model, provider, tier, IDs.
Env-only; no HTTP call. Useful so the agent can answer "what model
am I?" correctly instead of guessing from a stale system prompt
that the operator may have changed between boots.
Returns the identity as a JSON-encoded string (the dispatch contract
every MCP tool in this module follows). Tests that want to assert
individual fields can call ``_runtime_identity_payload()`` directly,
or ``json.loads`` the return value.
Always permitted by RBAC — there is no sensitive information here
that isn't already available to the process via ``os.environ``.
The point of the tool is to surface those env values to the agent
layer in a stable, documented shape rather than expecting every
agent runtime to know to ``echo $MODEL``.
"""
return json.dumps(_runtime_identity_payload(), indent=2)
async def tool_update_agent_card(card: Any) -> str:
"""Update this workspace's agent_card on the platform.
POSTs the provided card to ``/registry/update-card`` with the
workspace's own bearer token (same auth path as ``tool_commit_memory``
and ``tool_get_workspace_info``). The platform validates required
fields server-side, replaces the stored card, and broadcasts an
``agent_card_updated`` event so the canvas updates live.
Args:
card: A JSON-serialisable object (typically a dict) holding the
new card. The platform validates required fields server-side.
Returns:
JSON-encoded string. Body:
- ``{"success": true, "status": "updated"}`` on success;
- ``{"success": false, "error": "<msg>", "status_code": <int>}``
on platform error;
- ``{"success": false, "error": "<reason>"}`` on local validation
(non-dict card, missing WORKSPACE_ID, network error).
Permission gate: this tool requires the ``memory.write`` RBAC
capability — same gate as ``tool_commit_memory``. The check runs
inline rather than at the dispatcher layer to keep ``a2a_mcp_server``
permission-agnostic (the gate sits with the implementation, not the
transport). Read-only roles get a clear error string back instead
of a 403 from the platform.
We re-check ``isinstance(card, dict)`` here defensively rather than
trust the MCP schema validator alone — the schema only constrains
the transport, not the in-process call surface used by tests and
sibling modules.
"""
payload = await _update_agent_card_impl(card)
return json.dumps(payload, indent=2)
async def _update_agent_card_impl(card: Any) -> dict[str, Any]:
"""Dict-returning core of ``tool_update_agent_card``.
Split out so tests can assert against the raw dict shape (status
codes, error messages) without re-parsing JSON on every assertion.
The string-returning ``tool_update_agent_card`` is a thin wrapper
invoked by the MCP dispatcher.
"""
# RBAC: require memory.write permission. Same gate as
# tool_commit_memory (the agent already needs this capability to
# persist anything outbound). Read-only roles can still call
# get_runtime_identity / get_workspace_info to introspect — those
# are env-only / read-only and have no inline gate.
if not _check_memory_write_permission():
return {
"success": False,
"error": (
"RBAC — this workspace does not have the 'memory.write' "
"permission required to update the agent_card."
),
}
if not isinstance(card, dict):
return {
"success": False,
"error": "card must be a JSON object (dict)",
}
ws_id = os.environ.get("WORKSPACE_ID", "")
if not ws_id:
return {
"success": False,
"error": "WORKSPACE_ID env not set; cannot identify caller",
}
try:
async with httpx.AsyncClient(timeout=10.0) as client:
resp = await client.post(
f"{PLATFORM_URL}/registry/update-card",
json={"workspace_id": ws_id, "agent_card": card},
headers=_auth_headers_for_heartbeat(),
)
if resp.status_code == 200:
body: dict[str, Any] = {}
try:
body = resp.json()
except Exception:
pass
return {
"success": True,
"status": body.get("status", "updated"),
}
# Non-200 — surface what the platform returned.
error_msg = ""
try:
error_msg = resp.json().get("error", "") or resp.text
except Exception:
error_msg = resp.text
return {
"success": False,
"status_code": resp.status_code,
"error": error_msg,
}
except Exception as e:
return {"success": False, "error": f"network error: {e}"}
+52
View File
@@ -340,6 +340,16 @@ _CLI_A2A_COMMAND_KEYWORDS: dict[str, str | None] = {
"delegate_task_async": "delegate --async",
"check_task_status": "status",
"get_workspace_info": "info",
# `get_runtime_identity` + `update_agent_card` are MCP-first
# capabilities — the CLI subprocess interface doesn't expose them
# today. `get_runtime_identity` is env-only and an agent on a
# CLI-only runtime can already `echo $MODEL` etc, so there's no
# functional gap. `update_agent_card` requires a JSON object
# argument that wouldn't survive a positional-arg shell invocation
# cleanly. Mapped to None — flip to a keyword if a2a_cli grows
# `identity` / `card` subcommands in the future.
"get_runtime_identity": None,
"update_agent_card": None,
# `broadcast_message` is not exposed via the CLI subprocess interface
# today — it's an MCP-first capability. If a2a_cli grows a `broadcast`
# subcommand, map it here and the alignment test will gate the change.
@@ -589,6 +599,28 @@ def _sanitize_for_external(msg: str) -> str:
import re as _re
msg = _re.sub(r"(?i)(?:bearer|token|api[_-]?key|sk-)[ :=]+[A-Za-z0-9_/.-]{20,}", "[REDACTED]", msg)
# Bare provider key with NO separator after the prefix — a real
# `sk-ant-api03-…` / `sk-…` key uses `-` (not `[ :=]`) so the rule
# above misses it. Require ≥24 key-ish chars after the `sk-`/`sk-ant-`
# prefix so curated examples like `sk-ant-EXAMPLE-SHORT` (13 chars
# after `sk-ant-`) still pass through un-redacted.
msg = _re.sub(r"(?i)\bsk-(?:ant-)?[A-Za-z0-9_-]{24,}", "[REDACTED]", msg)
# JSON-quoted credential values: {"token": "…"} / {"apiKey": "…"} /
# {"secret": "…"} / {"password": "…"}. Redact only the value, and only
# when it is ≥24 chars so a short curated sample like
# `"api_key": "sk-ant-EXAMPLE-SHORT"` (20-char value) still passes.
msg = _re.sub(
r'(?i)("(?:token|api[_-]?key|secret|password)"\s*:\s*")[^"]{24,}(")',
r"\1[REDACTED]\2",
msg,
)
# AWS secret access key in `aws_secret_access_key=…` form (env dumps,
# boto tracebacks). The base64-ish value runs until whitespace/quote.
msg = _re.sub(
r"(?i)(aws_secret_access_key\s*[:=]\s*)\S+",
r"\1[REDACTED]",
msg,
)
# Absolute paths: /etc/shadow, /home/user/.aws/credentials, etc.
msg = _re.sub(r"(?:/[^/\s]+){2,}", lambda m: m.group(0) if len(m.group(0)) < 60 else "[REDACTED_PATH]", msg)
return msg
@@ -598,6 +630,7 @@ def sanitize_agent_error(
exc: BaseException | None = None,
category: str | None = None,
stderr: str | None = None,
reason: str | None = None,
) -> str:
"""Render an agent-side failure into a user-safe error message.
@@ -605,6 +638,18 @@ def sanitize_agent_error(
category string (e.g. from `classify_subprocess_error`). If both are
given, `category` wins. If neither, the tag defaults to "unknown".
When ``reason`` is provided (internal#211/#212), it is a *pre-curated,
user-actionable, secret-safe* explanation built by the caller from a
provider-side failure — e.g. a 403 "Your organization has disabled
Claude subscription access · Use an Anthropic API key instead, or ask
your admin to enable access" with error code ``oauth_org_not_allowed``.
This text is exactly what the user needs to self-serve, so it is
surfaced VERBATIM as the message instead of being collapsed to the
opaque exception class name. It still passes through the
key/token/bearer/path scrubber as a belt-and-braces second pass so a
buggy caller can't leak a credential that snuck into the reason.
``reason`` wins over ``stderr``; both lose to neither being set.
When ``stderr`` is provided (e.g. the first ~1 KB of a subprocess stderr
or HTTP error body), it is sanitized and appended to the output so the
A2A caller gets actionable context without needing to dig through workspace
@@ -619,6 +664,13 @@ def sanitize_agent_error(
else:
tag = "unknown"
if reason:
# Curated, user-actionable reason — surface it as the message.
# Still scrub: a 403/auth/quota message is safe, but the scrubber
# is cheap insurance against a caller that didn't curate cleanly.
clean = _sanitize_for_external(reason[:_MAX_STDERR_PREVIEW])
return f"Agent error ({tag}): {clean}"
if stderr:
# Truncate and sanitize before including — prevents DoS via
# a malicious or buggy peer injecting a huge error body, and
+59
View File
@@ -57,12 +57,14 @@ from a2a_tools import (
tool_commit_memory,
tool_delegate_task,
tool_delegate_task_async,
tool_get_runtime_identity,
tool_get_workspace_info,
tool_inbox_peek,
tool_inbox_pop,
tool_list_peers,
tool_recall_memory,
tool_send_message_to_user,
tool_update_agent_card,
tool_wait_for_message,
)
@@ -289,6 +291,61 @@ _GET_WORKSPACE_INFO = ToolSpec(
section=A2A_SECTION,
)
_GET_RUNTIME_IDENTITY = ToolSpec(
name="get_runtime_identity",
short=(
"Return this runtime's identity — model, model_provider, tier, "
"workspace_id, runtime template. Reads from process env; no HTTP call."
),
when_to_use=(
"Use this to answer 'what model am I?' truthfully instead of "
"guessing from a stale system prompt — the operator may have "
"routed you to a different model via persona env between boots. "
"Always permitted by RBAC: even read-only agents may know what "
"model they are. Distinct from get_workspace_info — that one "
"calls the platform for ID/role/tier/parent (workspace metadata); "
"this one returns the live process env (MODEL, MODEL_PROVIDER, "
"MOLECULE_MODEL, ANTHROPIC_BASE_URL, TIER, WORKSPACE_ID, "
"ADAPTER_MODULE)."
),
input_schema={"type": "object", "properties": {}},
impl=tool_get_runtime_identity,
section=A2A_SECTION,
)
_UPDATE_AGENT_CARD = ToolSpec(
name="update_agent_card",
short=(
"Replace this workspace's agent_card on the platform. The "
"platform validates required fields and broadcasts an "
"agent_card_updated event so the canvas reflects the change live."
),
when_to_use=(
"Use when the workspace's capabilities, skills, description, or "
"name change and the canvas display needs to follow. The "
"platform stores the new card and pushes an "
"``agent_card_updated`` event to subscribers. Gated behind the "
"``memory.write`` RBAC capability — read-only roles cannot "
"rewrite the card. Tier-1+ owners always have this capability."
),
input_schema={
"type": "object",
"properties": {
"card": {
"type": "object",
"description": (
"The new agent_card object (name, version, "
"description, skills, etc). Server-side validation "
"rejects payloads missing required fields."
),
},
},
"required": ["card"],
},
impl=tool_update_agent_card,
section=A2A_SECTION,
)
_BROADCAST_MESSAGE = ToolSpec(
name="broadcast_message",
short=(
@@ -642,6 +699,8 @@ TOOLS: list[ToolSpec] = [
_CHECK_TASK_STATUS,
_LIST_PEERS,
_GET_WORKSPACE_INFO,
_GET_RUNTIME_IDENTITY,
_UPDATE_AGENT_CARD,
_BROADCAST_MESSAGE,
_SEND_MESSAGE_TO_USER,
# Inbox (standalone-only; in-container returns informational error)
@@ -5,6 +5,8 @@
- **check_task_status**: Poll the status of a task started with delegate_task_async; returns result when done.
- **list_peers**: List the workspaces this agent can communicate with — name, ID, status, role for each.
- **get_workspace_info**: Get this workspace's own info — ID, name, role, tier, parent, status.
- **get_runtime_identity**: Return this runtime's identity — model, model_provider, tier, workspace_id, runtime template. Reads from process env; no HTTP call.
- **update_agent_card**: Replace this workspace's agent_card on the platform. The platform validates required fields and broadcasts an agent_card_updated event so the canvas reflects the change live.
- **broadcast_message**: Send a message to ALL agent workspaces in the org simultaneously. Requires broadcast_enabled=true on this workspace (set by user/admin).
- **send_message_to_user**: Send a message directly to the user's canvas chat — pushed instantly via WebSocket. Use this to: (1) acknowledge a task immediately ('Got it, I'll start working on this'), (2) send interim progress updates while doing long work, (3) deliver follow-up results after delegation completes, (4) attach files (zip, pdf, csv, image) for the user to download via the `attachments` field (NEVER paste file URLs in `message`). The message appears in the user's chat as if you're proactively reaching out.
- **wait_for_message**: Block until the next inbound message (canvas user OR peer agent) arrives, or until ``timeout_secs`` elapses.
@@ -27,6 +29,12 @@ Call this first when you need to delegate but don't know the target's ID. Access
### get_workspace_info
Use to introspect your own identity (e.g. before reporting back to the user, or to determine whether you're a tier-0 root that can write GLOBAL memory).
### get_runtime_identity
Use this to answer 'what model am I?' truthfully instead of guessing from a stale system prompt — the operator may have routed you to a different model via persona env between boots. Always permitted by RBAC: even read-only agents may know what model they are. Distinct from get_workspace_info — that one calls the platform for ID/role/tier/parent (workspace metadata); this one returns the live process env (MODEL, MODEL_PROVIDER, MOLECULE_MODEL, ANTHROPIC_BASE_URL, TIER, WORKSPACE_ID, ADAPTER_MODULE).
### update_agent_card
Use when the workspace's capabilities, skills, description, or name change and the canvas display needs to follow. The platform stores the new card and pushes an ``agent_card_updated`` event to subscribers. Gated behind the ``memory.write`` RBAC capability — read-only roles cannot rewrite the card. Tier-1+ owners always have this capability.
### broadcast_message
Use for urgent, org-wide signals: critical status changes, emergency stop instructions, coordinated task announcements. Every non-removed workspace receives the message in its activity log (poll-mode agents see it on their next poll; push-mode canvases get a real-time banner). This tool returns an error if broadcast_enabled is false — a user or admin must enable it via the workspace abilities settings first.
+390
View File
@@ -0,0 +1,390 @@
"""Tests for ``tool_get_runtime_identity`` and ``tool_update_agent_card``.
These two MCP tools close the T4-tier workspace owner-permission gaps
reported via the canvas:
- the agent could not update its own ``agent_card`` (no MCP tool
wrapped the existing ``POST /registry/update-card`` endpoint);
- the agent could not identify which model it was running (the
``MODEL`` env var is injected by ``provisioner.workspace_provision``
but nothing surfaced it back to the agent).
Ported from molecule-ai-workspace-runtime PR#17 (mirror-only repo;
canonical edit point per ``reference_runtime_repo_is_mirror_only``).
Adapted to core's conventions:
* tool functions return ``str`` (JSON-encoded), matching every other
tool in ``a2a_tools_*`` modules. Tests ``json.loads`` to inspect.
* permission check ``memory.write`` runs inline in
``tool_update_agent_card`` (same pattern as
``a2a_tools_memory.tool_commit_memory``).
* ``WORKSPACE_ID`` is read directly from ``os.environ`` — core does
not have the runtime's validated-cache layer (``molecule_runtime.
builtin_tools.validation``).
"""
from __future__ import annotations
import json
import pytest
# --- Drift gate: re-export aliases on a2a_tools ------------------------------
class TestBackCompatAliases:
"""Pin that ``a2a_tools.tool_*`` resolves to the same callable as
``a2a_tools_identity.tool_*``. Refactor wrapping (e.g. a doc-string
wrapper that loses the function identity) silently breaks call
sites that ``patch("a2a_tools.tool_update_agent_card", ...)`` —
this gate makes that drift fail fast."""
def test_tool_get_runtime_identity_alias(self):
import a2a_tools
import a2a_tools_identity
assert a2a_tools.tool_get_runtime_identity is a2a_tools_identity.tool_get_runtime_identity
def test_tool_update_agent_card_alias(self):
import a2a_tools
import a2a_tools_identity
assert a2a_tools.tool_update_agent_card is a2a_tools_identity.tool_update_agent_card
# --- tool_get_runtime_identity ----------------------------------------------
class TestGetRuntimeIdentity:
"""The tool returns env-derived runtime identity. No HTTP call."""
@pytest.mark.asyncio
async def test_returns_all_known_env_fields(self, monkeypatch):
from a2a_tools_identity import tool_get_runtime_identity
monkeypatch.setenv("MODEL", "claude-opus-4-7")
monkeypatch.setenv("MODEL_PROVIDER", "anthropic")
monkeypatch.setenv("TIER", "T4")
monkeypatch.setenv("WORKSPACE_ID", "ws-abc")
monkeypatch.setenv("ADAPTER_MODULE", "adapter")
monkeypatch.setenv("MOLECULE_MODEL", "claude-opus-4-7")
monkeypatch.setenv("ANTHROPIC_BASE_URL", "https://api.anthropic.com")
out = await tool_get_runtime_identity()
# MCP tools return JSON-encoded strings (matches the contract
# every other tool_* in a2a_tools_* uses).
assert isinstance(out, str)
parsed = json.loads(out)
assert parsed["model"] == "claude-opus-4-7"
assert parsed["model_provider"] == "anthropic"
assert parsed["tier"] == "T4"
assert parsed["workspace_id"] == "ws-abc"
assert parsed["runtime"] == "adapter"
assert parsed["molecule_model"] == "claude-opus-4-7"
assert parsed["anthropic_base_url"] == "https://api.anthropic.com"
@pytest.mark.asyncio
async def test_missing_env_returns_empty_strings(self, monkeypatch):
"""Tool MUST NOT raise when env vars are absent — every key is
present but the value is the empty string. The agent then knows
the slot exists but is unset."""
from a2a_tools_identity import tool_get_runtime_identity
for var in (
"MODEL", "MODEL_PROVIDER", "TIER", "WORKSPACE_ID",
"ADAPTER_MODULE", "MOLECULE_MODEL", "ANTHROPIC_BASE_URL",
):
monkeypatch.delenv(var, raising=False)
parsed = json.loads(await tool_get_runtime_identity())
assert parsed["model"] == ""
assert parsed["model_provider"] == ""
assert parsed["tier"] == ""
assert parsed["workspace_id"] == ""
assert parsed["runtime"] == ""
assert parsed["molecule_model"] == ""
assert parsed["anthropic_base_url"] == ""
@pytest.mark.asyncio
async def test_no_http_call_made(self, monkeypatch):
"""``get_runtime_identity`` is env-only — must not open
httpx.AsyncClient even if the call would otherwise succeed.
Tripwire any client construction."""
import httpx
from a2a_tools_identity import tool_get_runtime_identity
class _Tripwire:
def __init__(self, *_a, **_kw):
raise AssertionError(
"tool_get_runtime_identity must not open httpx.AsyncClient"
)
monkeypatch.setattr(httpx, "AsyncClient", _Tripwire)
# Must not raise.
await tool_get_runtime_identity()
@pytest.mark.asyncio
async def test_helper_dict_matches_string_payload(self, monkeypatch):
"""``_runtime_identity_payload`` is the dict-returning helper
used by both the public tool and tests. Verify the public tool
json.dumps the same dict — no field is dropped or renamed by
the encoding step."""
from a2a_tools_identity import (
_runtime_identity_payload,
tool_get_runtime_identity,
)
monkeypatch.setenv("MODEL", "claude-opus-4-7")
monkeypatch.setenv("TIER", "T4")
monkeypatch.setenv("WORKSPACE_ID", "ws-helper-check")
helper = _runtime_identity_payload()
tool_str = await tool_get_runtime_identity()
assert json.loads(tool_str) == helper
# --- tool_update_agent_card -------------------------------------------------
class _MockResponse:
def __init__(self, status_code: int, payload: dict):
self.status_code = status_code
self._payload = payload
self.text = json.dumps(payload)
def json(self):
return self._payload
class _MockClient:
"""Drop-in for httpx.AsyncClient context manager.
Records the URL + json body + headers the tool POSTed so the test
can assert against them. Returns the canned _MockResponse passed
in at construction time.
"""
def __init__(self, *, response: _MockResponse, captured: dict):
self._response = response
self._captured = captured
async def __aenter__(self):
return self
async def __aexit__(self, *_args):
return False
async def post(self, url, *, json=None, headers=None, **_kw): # noqa: A002
self._captured["url"] = url
self._captured["json"] = json
self._captured["headers"] = headers
return self._response
@pytest.fixture
def _grant_memory_write(monkeypatch):
"""Force the inline RBAC gate inside ``tool_update_agent_card`` to
succeed. The gate calls
``a2a_tools_rbac.check_memory_write_permission`` which inspects
``$MOLECULE_ROLES`` / the role table; the patch sidesteps that
machinery so tests can focus on the platform-call shape.
"""
import a2a_tools_identity
monkeypatch.setattr(
a2a_tools_identity, "_check_memory_write_permission", lambda: True
)
class TestUpdateAgentCard:
@pytest.mark.asyncio
async def test_posts_to_registry_update_card(
self, monkeypatch, _grant_memory_write,
):
"""Hits POST {PLATFORM_URL}/registry/update-card with the
workspace bearer and the {workspace_id, agent_card} body shape
the platform handler expects (workspace-server
``internal/handlers/registry.go``)."""
import a2a_tools_identity
monkeypatch.setenv("WORKSPACE_ID", "ws-42")
# Ensure PLATFORM_URL re-import sees a deterministic value —
# a2a_client imports it at module load so we patch the symbol
# on a2a_tools_identity directly (the module's own reference).
monkeypatch.setattr(a2a_tools_identity, "PLATFORM_URL", "http://test.invalid")
captured: dict = {}
response = _MockResponse(200, {"status": "updated"})
def _client_factory(*_a, **_kw):
return _MockClient(response=response, captured=captured)
monkeypatch.setattr(a2a_tools_identity.httpx, "AsyncClient", _client_factory)
monkeypatch.setattr(
a2a_tools_identity, "_auth_headers_for_heartbeat",
lambda: {"Authorization": "Bearer ws-token-xyz"},
)
card = {"name": "agent-foo", "version": "0.1.0", "description": "demo"}
result_str = await a2a_tools_identity.tool_update_agent_card(card)
result = json.loads(result_str)
# URL: PLATFORM_URL + /registry/update-card
assert captured["url"] == "http://test.invalid/registry/update-card"
# The platform handler expects {workspace_id, agent_card}; the
# agent_card is the raw object the agent submitted.
body = captured["json"]
assert body["workspace_id"] == "ws-42"
assert body["agent_card"] == card
# Auth header from auth_headers_for_heartbeat is forwarded
# verbatim — same path commit_memory uses.
assert captured["headers"]["Authorization"] == "Bearer ws-token-xyz"
assert result["success"] is True
assert result["status"] == "updated"
@pytest.mark.asyncio
async def test_propagates_server_error(
self, monkeypatch, _grant_memory_write,
):
"""Non-200 from platform surfaces as a structured error to the
agent. The agent sees {success:false, status_code, error} and
can decide whether to retry, fall back, or escalate."""
import a2a_tools_identity
monkeypatch.setenv("WORKSPACE_ID", "ws-42")
monkeypatch.setattr(a2a_tools_identity, "PLATFORM_URL", "http://test.invalid")
captured: dict = {}
response = _MockResponse(400, {"error": "invalid card"})
monkeypatch.setattr(
a2a_tools_identity.httpx, "AsyncClient",
lambda *a, **kw: _MockClient(response=response, captured=captured),
)
monkeypatch.setattr(
a2a_tools_identity, "_auth_headers_for_heartbeat", lambda: {},
)
result = json.loads(
await a2a_tools_identity.tool_update_agent_card({"name": "x"})
)
assert result["success"] is False
assert result["status_code"] == 400
assert "invalid card" in str(result["error"]).lower()
@pytest.mark.asyncio
async def test_rejects_non_dict_card(self, _grant_memory_write):
"""The MCP schema constrains transport callers to pass a dict;
in-process callers (tests, sibling modules) can still pass any
type. Reject non-dict defensively so the platform isn't asked
to validate JSON-encoded strings or lists."""
from a2a_tools_identity import tool_update_agent_card
result = json.loads(await tool_update_agent_card("not-a-dict"))
assert result["success"] is False
assert "dict" in str(result["error"]).lower()
@pytest.mark.asyncio
async def test_workspace_id_missing_returns_error(
self, monkeypatch, _grant_memory_write,
):
"""If WORKSPACE_ID is not set the tool refuses to issue the
request — it would otherwise POST with an empty workspace_id
and let the platform return a confusing 400."""
from a2a_tools_identity import tool_update_agent_card
monkeypatch.delenv("WORKSPACE_ID", raising=False)
result = json.loads(await tool_update_agent_card({"name": "x"}))
assert result["success"] is False
assert "workspace_id" in str(result["error"]).lower()
@pytest.mark.asyncio
async def test_denies_when_memory_write_permission_missing(self, monkeypatch):
"""The agent's RBAC role must grant ``memory.write`` to update
the card. Read-only roles get an RBAC error string back
immediately, never touching the platform."""
import a2a_tools_identity
monkeypatch.setenv("WORKSPACE_ID", "ws-42")
monkeypatch.setattr(
a2a_tools_identity, "_check_memory_write_permission", lambda: False,
)
# Tripwire httpx — must not be called when RBAC denies.
import httpx
class _Tripwire:
def __init__(self, *_a, **_kw):
raise AssertionError("RBAC denial must short-circuit before httpx call")
monkeypatch.setattr(httpx, "AsyncClient", _Tripwire)
result = json.loads(
await a2a_tools_identity.tool_update_agent_card({"name": "x"}),
)
assert result["success"] is False
assert "memory.write" in str(result["error"]).lower()
@pytest.mark.asyncio
async def test_network_exception_returns_structured_error(
self, monkeypatch, _grant_memory_write,
):
"""A network exception (DNS failure, connect timeout, etc) is
wrapped into a structured error dict instead of bubbling up
to the MCP transport layer."""
import a2a_tools_identity
monkeypatch.setenv("WORKSPACE_ID", "ws-42")
monkeypatch.setattr(a2a_tools_identity, "PLATFORM_URL", "http://test.invalid")
class _ExplodingClient:
async def __aenter__(self):
return self
async def __aexit__(self, *_a):
return False
async def post(self, *_a, **_kw):
raise RuntimeError("simulated DNS failure")
monkeypatch.setattr(
a2a_tools_identity.httpx, "AsyncClient",
lambda *a, **kw: _ExplodingClient(),
)
result = json.loads(
await a2a_tools_identity.tool_update_agent_card({"name": "x"})
)
assert result["success"] is False
assert "network" in str(result["error"]).lower()
# --- Registry contract ------------------------------------------------------
class TestRegistryContract:
"""Pin the new tools' registration in platform_tools.registry. The
structural tests in ``test_platform_tools.py`` already check
registry↔MCP alignment; these are tighter assertions specific to
the two new tools so a future contributor deleting one entry sees
a focused failure."""
def test_get_runtime_identity_in_registry(self):
from platform_tools.registry import by_name
spec = by_name("get_runtime_identity")
assert spec.section == "a2a"
# No input parameters — env-only call.
assert spec.input_schema == {"type": "object", "properties": {}}
# impl points at the actual tool function, not a shim.
from a2a_tools_identity import tool_get_runtime_identity
assert spec.impl is tool_get_runtime_identity
def test_update_agent_card_in_registry(self):
from platform_tools.registry import by_name
spec = by_name("update_agent_card")
assert spec.section == "a2a"
assert "card" in spec.input_schema["properties"]
assert spec.input_schema["required"] == ["card"]
from a2a_tools_identity import tool_update_agent_card
assert spec.impl is tool_update_agent_card
+117
View File
@@ -788,6 +788,123 @@ def test_sanitize_agent_error_stderr_combined_with_existing_tests():
assert "workspace logs" in out
# ─── reason passthrough (internal#211/#212: surface actionable provider error) ───
def test_sanitize_agent_error_reason_surfaced_verbatim():
"""A curated provider reason is shown to the user, not collapsed to the
exception class name. This is the internal#211 regression: a 403
org-disabled message must reach the canvas."""
reason = (
"provider HTTP 403 — oauth_org_not_allowed — Your organization has "
"disabled Claude subscription access for Claude Code · Use an "
"Anthropic API key instead, or ask your admin to enable access"
)
class _ResultErr(Exception):
pass
out = sanitize_agent_error(exc=_ResultErr("opaque"), reason=reason)
# The actionable provider guidance and status code must be visible.
assert "403" in out
assert "oauth_org_not_allowed" in out
assert "disabled Claude subscription access" in out
assert "ask your admin to enable access" in out
# NOT the old opaque form.
assert "see workspace logs" not in out
def test_sanitize_agent_error_reason_still_scrubs_secrets():
"""Even on the reason path the key/token scrubber runs — a buggy caller
that lets a bearer token into the reason still gets it redacted."""
leaky = (
"provider HTTP 401 — auth failed — Authorization: Bearer "
"PLACEHOLDER_LONG_TOKEN_0123456789abcdefghijklm please re-auth"
)
out = sanitize_agent_error(reason=leaky)
assert "[REDACTED]" in out
assert "PLACEHOLDER_LONG_TOKEN_0123456789abcdefghijklm" not in out
# The non-secret guidance still survives the scrub.
assert "401" in out
assert "please re-auth" in out
def test_sanitize_agent_error_reason_scrubs_all_secret_formats():
"""The scrubber must redact every realistic credential shape — not just
the `Bearer <tok>` form the original test happened to exercise
(internal#212 review finding: bare `sk-ant-api03-…` keys, JSON-quoted
"token"/"apiKey" values, and `aws_secret_access_key=` all leaked).
All curated/actionable guidance must still survive the scrub.
"""
# 1. Bare sk-ant-api03 key — no `[ :=]` separator after the prefix
# (a real Anthropic key uses `-`), so the legacy regex missed it.
bare = (
"provider HTTP 401 — auth failed — invalid key "
"sk-FAKEPLACEHOLDERabcdefghijklmnopqrstuvwxy0123456789 "
"please re-auth"
)
out = sanitize_agent_error(reason=bare)
assert "sk-FAKEPLACEHOLDERabcdefghijklmnopqrstuvwxy0123456789" not in out
assert "[REDACTED]" in out
assert "401" in out # actionable status survives
assert "please re-auth" in out # actionable guidance survives
# 2. JSON-quoted "token" / "apiKey" values.
jblob = (
'provider error — config dump {"token": '
'"abcDEF0123456789ghIJKL0123456789mnopQRST", "apiKey": '
'"anon_fakefakefakefakefakefakefakefakefakefake"} — '
"use an API key instead"
)
out = sanitize_agent_error(reason=jblob)
assert "abcDEF0123456789ghIJKL0123456789mnopQRST" not in out
assert "anon_fakefakefakefakefakefakefakefakefakefake" not in out
assert "[REDACTED]" in out
assert "use an API key instead" in out # actionable guidance survives
# 3. aws_secret_access_key=… form.
awsblob = (
"provider HTTP 403 — boto credential error "
"aws_secret_access_key=wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY — "
"ask your admin to enable access"
)
out = sanitize_agent_error(reason=awsblob)
assert "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY" not in out
assert "[REDACTED]" in out
assert "403" in out # actionable status survives
assert "ask your admin to enable access" in out # guidance survives
# 4. Regression: the original Bearer form still redacts.
# Uses PLACEHOLDER_LONG_TOKEN (>=40 chars, no sk-ant- prefix) to avoid
# triggering the secret-scan workflow pattern
# `sk-ant-[A-Za-z0-9_-]{40,}`.
bearer = (
"provider HTTP 401 — Authorization: Bearer "
"PLACEHOLDER_LONG_TOKEN_9876543210abcdefghij re-auth"
)
out = sanitize_agent_error(reason=bearer)
assert "PLACEHOLDER_LONG_TOKEN_9876543210abcdefghij" not in out
assert "[REDACTED]" in out
assert "re-auth" in out
def test_sanitize_agent_error_reason_wins_over_stderr():
"""When both reason and stderr are passed, the curated reason wins."""
out = sanitize_agent_error(
reason="provider HTTP 403 — use an API key",
stderr="raw subprocess noise that should not be shown",
)
assert "use an API key" in out
assert "raw subprocess noise" not in out
def test_sanitize_agent_error_no_reason_unchanged():
"""Omitting reason preserves the original generic behavior."""
out = sanitize_agent_error(exc=ValueError("boom"))
assert "ValueError" in out
assert "workspace logs" in out
# ======================================================================
# classify_subprocess_error