Security Auditor findings on PR #611:
Fix 1 (BLOCKING): Remove budget_limit handling from Update() entirely.
PATCH /workspaces/:id uses ValidateAnyToken — any enrolled workspace bearer
could self-clear its own spending ceiling. The dedicated AdminAuth-gated
PATCH /workspaces/:id/budget is the only authorised write path.
Fix 2 (MEDIUM): Strip budget_limit and monthly_spend from Get() response
before c.JSON(). GET /workspaces/:id is on the open router — any caller
with a valid UUID must not read billing data.
Also updates four existing tests in workspace_budget_test.go that encoded
the old (insecure) behaviour, and adds three new regression tests.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Workspace agents could previously call PATCH /workspaces/:id/budget with their
own bearer token and set budget_limit=null, defeating the entire spend enforcement
feature. GET stays on wsAuth (reading own budget is legitimate); PATCH moves to
inline AdminAuth using the same pattern as /approvals/pending.
No existing tests needed updating — all budget PATCH tests call the handler
directly and are unaffected by router-level middleware changes.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Budget limit exceeded on A2A proxy now returns HTTP 402 PaymentRequired
instead of 429 TooManyRequests, matching the issue spec and the FE amber
banner check. Updates a2a_proxy.go, workspace_budget_test.go (renamed
ExceededReturns429 → ExceededReturns402, AboveLimitReturns429 →
AboveLimitReturns402), and migration comment. All go test ./... pass.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- New BudgetHandler with GetBudget and PatchBudget methods
- GET returns budget_limit (null or int64 USD cents), monthly_spend,
and computed budget_remaining (null when no limit, can be negative
when over-budget so callers can see the magnitude of the overage)
- PATCH accepts {budget_limit: int64|null}; null clears the ceiling;
validates non-negative values; re-reads DB to echo final state
- Both handlers are wired in router.go under the WorkspaceAuth group
- 14 unit tests covering happy paths, 404, 400 validation, DB errors,
over-budget state, zero limit, and clear-limit round-trip
- All 20 packages pass go test ./... and go build ./... is clean
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Migration 025: ADD COLUMN budget_limit BIGINT DEFAULT NULL and
monthly_spend BIGINT NOT NULL DEFAULT 0 to workspaces table
- Models: BudgetLimit *int64 in CreateWorkspacePayload;
MonthlySpend int64 in HeartbeatPayload
- workspace.go: scanWorkspaceRow, workspaceListQuery, Get, Create, and
Update all handle budget_limit/monthly_spend; budget_limit is gated
as a sensitiveUpdateField
- registry.go: heartbeat conditionally writes monthly_spend only when
payload.MonthlySpend > 0 (avoids overwriting with zero)
- a2a_proxy.go: checkWorkspaceBudget() returns 429 when
monthly_spend >= budget_limit (NULL = no limit; fail-open on DB error)
- Tests: 8 new workspace_budget_test.go tests + patched existing tests
for the 20-column scanWorkspaceRow and 10-param CREATE INSERT
Field type: BIGINT (int64), units: USD cents (budget_limit=500 = $5.00/month)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Three add/add + content conflicts, all mechanical:
- WorkspaceUsage.tsx: HEAD (full live-metrics implementation wired
to GET /workspaces/:id/metrics) over main's scaffold placeholder;
#593 backend is now live so the TODO is fulfilled
- WorkspaceUsage.test.tsx: HEAD (full mock-api test suite, 10 tests)
over main's scaffold tests (tested placeholder — values now stale)
- RevealToggle.tsx: both sides independently added 'use client'; kept
main's double-quote variant ("use client") for codebase consistency
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
A malicious or buggy agent could report MonthlySpend = math.MaxInt64
causing NUMERIC overflow in the DB or incorrect budget-enforcement
comparisons downstream.
Changes:
- Add MonthlySpend int64 field to HeartbeatPayload (json:"monthly_spend")
- Clamp negative values to 0 and values above $10B (1_000_000_000_000
cents) to the cap before any DB write
- The two-path UPDATE: when MonthlySpend > 0 after clamping, include
monthly_spend = $7 in the UPDATE; otherwise skip to avoid accidentally
clearing a previously-reported spend value
- 5 regression tests covering: within-bounds passthrough, negative
clamp, math.MaxInt64 overflow clamp, exact-cap boundary, and
zero/omitted no-update path
Note: this branch introduces MonthlySpend to HeartbeatPayload; it will
need trivial conflict resolution when feat/issue-541-budget-limit-backend
merges, as that branch also adds the field (without the cap). Keep this
branch's clamping logic.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Merge gate passed (all 7 gates). Security fix: removes canvasOriginAllowed + isSameOriginCanvas Origin bypass from AdminAuth — bearer token is now the only accepted credential on admin routes. 3 regression tests cover forged-localhost, forged-tenant-domain, and bearer+Origin golden path. Auth PR — CEO explicit approval confirmed in chat. UNSTABLE = known GitHub App token scope gap.
The vi.mock("../../../store/canvas") call was nested inside an it()
block. Vitest hoists all vi.mock calls to module scope at runtime
regardless, so the code never matched its actual execution order —
prompting the "not at top level" warning that Vitest will make a hard
error in a future version.
Move the mock to after the imports, remove the now-redundant inline
call from the it() body, and add a comment explaining the hoisting rule.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
parseInt("0", 10) || null evaluates to null, silently converting a
zero-credit budget to unlimited. Switch to raw !== "" ? parseInt() : null
so budget_limit: 0 is sent correctly. Adds regression test.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds useId() to the InputField helper in CreateWorkspaceDialog so every
<label> is wired to its <input> via htmlFor/id. Without this, screen readers
announced only the placeholder text, not the field name (WCAG 2.1 SC 1.3.1
Level A violation, build 4JIwTGVMjDGNLO8iMGJeC).
Affected fields: Name (required), Role, Budget limit (USD), Template.
The Hermes provider fields were already correctly wired.
Adds 6 new tests in CreateWorkspaceDialog.a11y.test.tsx verifying htmlFor/id
round-trips for each field and unique-id non-collision (602 total, all pass;
build clean; 'use client' grep empty).
Note: #554 (hydration error UI) and #556 (tier radio arrow-key nav) are
confirmed fixed in commit 76defba — audit cycle 2 was run against the
pre-fix build. #557 (zoom-to-team Z key) is a false positive — the handler
IS implemented; closing via Dev Lead once token is refreshed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds a dedicated BudgetSection component to the workspace details panel:
- GET /workspaces/:id/budget on mount — populates live stats (used/limit/remaining)
- Stats row + blue-500 progress bar (capped at 100%; hidden when unlimited)
- PATCH /workspaces/:id/budget for saving; input blank → budget_limit: null
- "Budget exceeded — messages blocked" amber/zinc-950 banner on any 402 response
(GET or PATCH); banner clears on a successful subsequent save
- 'use client'; dark zinc theme throughout (zinc-800/700 inputs, blue-500 accents)
DetailsTab refactored: inline budget_limit fields removed; BudgetSection mounted
as a self-contained section between Workspace and Skills. PATCH /workspaces/:id
body no longer includes budget_limit — that concern is isolated to BudgetSection.
Tests: 21 new cases in BudgetSection.test.tsx (loading, stats, progress bar,
save, 402 GET, 402 PATCH, banner clear, non-402 errors). BudgetLimit.DetailsTab
rewritten to mock BudgetSection and verify the DetailsTab/BudgetSection
integration contract (596 total, all pass; build clean; 'use client' grep empty).
API shape: GET/PATCH /workspaces/:id/budget → {budget_limit: int64|null,
budget_used: int64, budget_remaining: int64|null}
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Adds optional Budget limit (USD) numeric field to CreateWorkspaceDialog;
blank = null (unlimited), populated = parsed float sent as budget_limit in
POST /workspaces body
- Adds budget_limit field to DetailsTab edit form; saves via
PATCH /workspaces/:id; pre-fills from current WorkspaceNodeData
- Shows 'Budget limit exceeded' warning badge when budgetUsed > budgetLimit
(forward-compatible — badge hidden when budgetUsed is absent)
- Extends WorkspaceData, WorkspaceNodeData, and buildNodesAndEdges to carry
budgetLimit / budgetUsed fields ready for backend hydration (issue #541 BE PR)
- Ships 22 new tests across CreateWorkspaceDialog and BudgetLimit.DetailsTab
suites (575 total, all passing); npm run build clean; 'use client' grep empty
API shape confirmed from workspace.go and CreateWorkspacePayload struct:
field name: budget_limit | type: number | null | units: USD
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adversarial or buggy agents can report INT64_MAX token counts via A2A
responses. Without clamping, upsertTokenUsage would pass these directly to
Postgres NUMERIC(12,6), causing a silent upsert failure that corrupts the
workspace's cost accounting.
Fix: clamp input_tokens/output_tokens to [0, 10_000_000] before any
arithmetic or DB write. 10M tokens/call is well above any real LLM API
response; clamped values still produce valid cost rows.
Adds 4 regression tests:
- TestUpsertTokenUsage_615_CapsInt64Max — INT64_MAX → maxTokensPerCall
- TestUpsertTokenUsage_615_CapsNegative — negative → 0 (no DB call)
- TestUpsertTokenUsage_615_NormalValuesUnchanged — passthrough for normal counts
- TestUpsertTokenUsage_615_ExactlyAtCap — at-cap value accepted unchanged
Closes#615
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
SecurityHeaders() middleware already sets X-Content-Type-Options: nosniff and
X-Frame-Options: DENY globally on every response (issue #151 / PR ~securityheaders).
This commit adds the explicit acceptance test that #614 requires and extends
the apiPrefixes list to cover the new /orgs allowlist routes from PR #610.
Changes:
- securityheaders.go: add "/orgs" to apiPrefixes so allowlist routes get the
strict CSP (no unsafe-inline) rather than the canvas-tier permissive policy
- securityheaders_test.go: TestSecurityHeaders_614_NosniffOnSSEAndAPIEndpoints
verifies the header is present on SSE endpoint, /settings/secrets, /events,
and /orgs paths; TestIsAPIPath gains /orgs cases
Closes#614
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The Origin header is trivially forgeable by any container on the Docker
network. Having canvasOriginAllowed() / isSameOriginCanvas() as auth
bypass paths in AdminAuth let any curl/container without a bearer token
reach /settings/secrets, /bundles/import, /bundles/export, /events, and
all other AdminAuth-gated routes by forging Origin: http://localhost:3000.
Fix: remove both Origin bypass branches from AdminAuth. Bearer token is
now the only accepted credential. Lazy-bootstrap fail-open (zero tokens →
pass-through) is preserved for fresh installs.
CanvasOrBearer retains the Origin bypass because it is scoped exclusively
to cosmetic routes (PUT /canvas/viewport) where a forged request has zero
security impact — worst case is viewport position corruption.
Added 3 regression tests:
- TestAdminAuth_623_ForgedOrigin_Returns401
- TestAdminAuth_623_ForgedCORSOrigin_Returns401
- TestAdminAuth_623_ValidBearer_WithOrigin_Passes
Closes#623, Closes#626
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
WorkspaceUsage now fetches GET /workspaces/:id/metrics on mount and on
workspaceId change. Displays input_tokens and output_tokens formatted
with toLocaleString, and estimated_cost_usd as $X.XXXXXX. Shows three
zinc-700 skeleton rows while loading; surfaces error text on failure.
Stale-fetch guard via ignore flag prevents state updates after unmount.
Also fixes missing 'use client' in RevealToggle.tsx (#603) — the
onClick handler requires client-side hydration.
Tests updated: 12 tests covering loading skeleton, API call correctness,
token formatting, cost formatting, error state, and workspaceId refetch.
All 551 canvas tests pass; build clean.
Closes#592Closes#603
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Add an org-scoped allowlist table so org admins can restrict which plugins
workspace agents are allowed to install. An empty allowlist means
allow-all (backward-compatible with existing deployments).
• migrations/027_org_plugin_allowlist.{up,down}.sql — new table + unique
index on (org_id, plugin_name)
• handlers/org_plugin_allowlist.go — resolveOrgID, checkOrgPluginAllowlist
(fail-open on DB errors), GetAllowlist, PutAllowlist (atomic tx replace)
• handlers/org_plugin_allowlist_test.go — 23 unit tests covering all
handler paths, resolveOrgID, and all checkOrgPluginAllowlist branches
• handlers/plugins_install.go — allowlist gate between resolveAndStage and
deliverToContainer; returns 403 if plugin is blocked
• router/router.go — GET/PUT /orgs/:id/plugins/allowlist under AdminAuth
All tests pass; go build ./... clean; gosec Issues: 0
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Migration 026 adds workspace_token_usage table (uuid pk, workspace_id FK with
CASCADE, period_start TIMESTAMPTZ, input_tokens, output_tokens, call_count,
estimated_cost_usd NUMERIC(12,6), updated_at) with a UNIQUE index on
(workspace_id, period_start) for day-granularity upserts.
A2A proxy (proxyA2ARequest) now spawns a detached goroutine after each
successful call to extractAndUpsertTokenUsage, which:
1. Parses usage.input_tokens / usage.output_tokens from result.usage
(JSON-RPC wrapper) with fallback to top-level usage (direct Anthropic).
2. Calls upsertTokenUsage — INSERT ... ON CONFLICT DO UPDATE so multi-
call days accumulate correctly. Estimated cost = input×$0.000003 +
output×$0.000015 (Claude Sonnet default; adjustable in a later phase).
Token tracking never blocks the critical A2A path.
New endpoint: GET /workspaces/:id/metrics (wsAuth — WorkspaceAuth bearer
bound to :id). Returns:
{"input_tokens":N,"output_tokens":N,"total_calls":N,
"estimated_cost_usd":"0.000000","period_start":"...","period_end":"..."}
404 if workspace missing. Period is current UTC day.
11 new tests (4 handler + 7 parse-unit); 19/19 packages pass.
Closes#593
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Unconditional list.extend() on repeated plugin install caused every
hook handler to be appended on each reinstall, leading to 3-4x duplicate
firings per event (PreToolUse, PostToolUse, Stop, etc.).
Fix: before appending each incoming handler, compute a fingerprint of
(matcher, frozenset-of-commands). Skip append if the fingerprint is
already present in the merged list. First-time installs are unaffected —
new handlers still land correctly.
Adds 7 unit tests covering: first install, double install, triple install,
different-matcher co-existence, different-command co-existence, existing
user hook preservation, and top-level key merge semantics.
Closes#566
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds WorkspaceUsage component to canvas/src/components/ with three
placeholder stat rows (Input tokens, Output tokens, Estimated cost)
and a "pending #593" badge. Wires into DetailsTab between the Workspace
and Skills sections. No API calls yet — fetch logic will be added once
GET /workspaces/:id/metrics lands in #593.
9 tests in WorkspaceUsage.test.tsx; all 548 canvas tests pass.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Add in-process SSE subscription mechanism to Broadcaster (SubscribeSSE,
deliverToSSE) so both RecordAndBroadcast *and* BroadcastOnly fan out to
SSE subscribers — critical because BroadcastOnly skips Redis pub/sub and
would be invisible to a Redis-only subscriber (AGENT_MESSAGE, A2A_RESPONSE,
TASK_UPDATED are all BroadcastOnly events).
- Add handlers/sse.go: SSEHandler.StreamEvents sets text/event-stream headers,
checks workspace existence (404 if missing), subscribes via broadcaster, and
wraps each WSMessage in an AG-UI envelope:
data: {"type":"<event>","timestamp":<unix_ms>,"data":{...}}\n\n
- Register wsAuth.GET("/workspaces/:id/events/stream") behind existing
WorkspaceAuth middleware — bearer token bound to :id.
- Add 6 tests: Content-Type, initial ping, AG-UI format, workspace filter
(cross-workspace events not leaked), 404 on missing workspace, multiple
sequential events.
All 19 packages pass. Build clean.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
MAF v1.0 shipped April 7 with multi-agent orchestration, native A2A+MCP,
AG-UI SSE protocol for streaming events to frontends. AG-UI is a direct
competitor to our WebSocket canvas. Added actionable gaps: AG-UI endpoint,
tool governance registry, cost transparency.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Hermes requires OPENROUTER_API_KEY (or any of its 15 providers).
Gemini CLI requires GOOGLE_API_KEY. Without these entries, the
MissingKeysModal doesn't fire and workspaces start without keys,
causing crash loops.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1. ScheduleTab + ChannelsTab: wrap toggle/delete in try/catch with
error feedback (was silently swallowing API failures)
2. MemoryTab: "+Add" button now auto-expands Advanced section
3. SidePanel: keyboard-navigated tabs scroll into view
4. TracesTab: emoji aria-hidden, env-var hint in <details>
5. page.tsx: show Spinner while hydrating instead of flash of EmptyState
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Standalone operational tool — doesn't belong in the platform core.
Should live in its own repo if needed.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Two new LOW-tier entries:
- dimos (dimensionalOS/dimos, 2.9k⭐, v0.0.11, MIT) — agentic OS for
robotics; MCP as primary agent interface; module/blueprint architecture
with typed stream passing; spatio-temporal RAG memory; hardware:
Unitree/AgileX/DJI/MAVLink. Watch for A2A support.
- Cloudflare Workers AI (Agents Week 2026) — unified inference layer:
70+ models, 14+ providers, auto-failover, streaming resilience, 330
global PoPs. Part of Cloudflare full-stack agent platform (+ Durable
Objects + Artifacts + Agents SDK + AI Search). Separate from previously
tracked Cloudflare Artifacts entry. Escalate to MEDIUM if Agents SDK
integrates all four primitives into one-click multi-agent deployment.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Without middleware, any caller on a non-production instance could mint a
bearer token for any workspace UUID with no authentication. AdminAuth is
defence-in-depth: on a fresh install (no tokens yet) it is fail-open so
the bootstrap path still works; once the first workspace enrolls a token
all callers must present a valid bearer.
Adds two router-level tests confirming the gate:
- TestTestTokenRoute_RequiresAdminAuth_WhenTokensExist → 401 with no header
- TestTestTokenRoute_FailOpenOnFreshInstall → 200 (bootstrap path intact)
Env-var gating inside GetTestToken is retained as a second layer.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
TR research (2026-04-17) confirmed v0.8/v0.9 do not exist in the A2A spec
history. Both Molecule AI (a2a-sdk==0.3.25) and CrewAI (protocol_version
default "0.3.0") are on spec v0.3.0 — zero-shim interop confirmed today.
Real future risk: A2A v1.0.0 (Mar 12 2026) — breaking changes in wire
format, agent card schema, OAuth flow. Neither side has migrated; shared
upgrade clock. Schedule coordinated migration before either upgrades.
Updates:
- YAML notable_changes: replace "v0.8/v0.9" with "v0.3.0, matches
a2a-sdk==0.3.25, zero-shim interop confirmed, v1.0.0 shared clock"
- Narrative: add A2A interop confirmed section + updated signals
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
