molecule-ai/molecule-ai-workspace-runtime
38
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
104 Commits 17 Branches 13 Tags 557 KiB
Python 96.7%
Shell 3.3%
ac8108a1a7
Go to file
Molecule AI Infra-Runtime-BE ac8108a1a7
Some checks failed
ci / mirror-guard (pull_request) Failing after 9s
Details
Secret scan / Scan diff for credential-shaped strings (pull_request) Successful in 9s
Details
fix(executor): sanitize peer delegation content in read_delegation_results (OFFSEC-003) 
peer-supplied `summary` and `response_preview` fields written to
DELEGATION_RESULTS_FILE by the heartbeat loop were injected into the
agent prompt without sanitization — a direct OFFSEC-003 injection path.

New `_detect_injection_safe()` helper wraps
`builtin_tools.compliance.detect_prompt_injection()` with lazy import
and fail-open behaviour. When injection patterns are detected in either
`summary` or `response_preview`, the field is replaced with "" before
formatting. The delegation metadata (status, task line) is preserved so
the agent still knows a delegation completed; only the malicious content
is stripped.

Fail-open: if builtin_tools.compliance is unavailable (e.g. minimal
test environment), the function logs a warning and passes text through.
This is acceptable because builtin_tools is always present in production
containers; the fail-open only affects degenerate test environments.

6 new tests covering: clean pass-through, injection in summary,
injection in preview, truncation of clean preview, no-file path,
fail-open when compliance unavailable.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-11 03:38:14 +00:00
.gitea/workflows fix(ci): allow mirror-guard bypass for CI infrastructure migration PRs 2026-05-11 03:09:49 +00:00
molecule_runtime fix(executor): sanitize peer delegation content in read_delegation_results (OFFSEC-003) 2026-05-11 03:38:14 +00:00
tests fix(executor): sanitize peer delegation content in read_delegation_results (OFFSEC-003) 2026-05-11 03:38:14 +00:00
.gitignore chore: gitignore credentials for molecule-ai-workspace-runtime 2026-04-16 09:18:48 -07:00
CONTRIBUTING.md fix(post-suspension): migrate github.com/Molecule-AI refs to git.moleculesai.app (Class G #168) 2026-05-07 13:02:35 -07:00
pyproject.toml fix(deps): enforce a2a-sdk>=1.0.0 to match code comment and KI-009 migration 2026-05-10 07:05:06 +00:00
README.md fix(post-suspension): migrate github.com/Molecule-AI refs to git.moleculesai.app (Class G #168) 2026-05-07 13:02:35 -07:00

README.md

molecule-ai-workspace-runtime

⚠️ This repo is a publish artifact, not the source of truth.

Runtime code lives in Molecule-AI/molecule-core → workspace/. This repo is regenerated and republished from there by the publish-runtime workflow on every runtime-v* tag.

Don't edit files here directly. PRs against this repo will not be merged. Open them against molecule-core instead.

Shared Python runtime infrastructure for all Molecule AI agent adapters.

This package provides the core machinery every Molecule AI workspace container needs:

  • A2A server — registers with the platform, heartbeats, serves A2A JSON-RPC
  • Adapter interface — BaseAdapter / AdapterConfig / SetupResult
  • Built-in tools — delegation, memory, approvals, sandbox, telemetry
  • Skill loader — loads and hot-reloads skill modules from /configs/skills/
  • Plugin system — per-workspace + shared plugin discovery and install
  • Config / preflight — YAML config loading with validation

Installation

pip install molecule-ai-workspace-runtime

Adapter discovery

The runtime discovers adapters in two ways:

  1. ADAPTER_MODULE env var (standalone adapter repos):

    ADAPTER_MODULE=adapter molecule-runtime

    The runtime imports adapter and calls adapter.Adapter.

  2. Subdirectory scan (monorepo local dev): falls back to scanning molecule_runtime/adapters/<runtime>/ and importing the matching subdir's Adapter class.

Contributing

Don't open PRs here. Send your change to Molecule-AI/molecule-core under the workspace/ directory. After your PR merges to main and a runtime-v* tag is pushed, the publish-runtime workflow rebuilds this mirror + uploads the new wheel to PyPI.

See docs/workspace-runtime-package.md for the full publishing flow.

Why this split

The runtime needs to ship as a PyPI artifact (so the 8 workspace template images can pip install it), but it also needs to evolve in lock-step with the platform's wire protocol (queue shape, A2A metadata, event payloads). A monorepo edit + auto-publish pipeline gives both: atomic cross-cutting changes, plus a clean PyPI release on every tag.

For the back-history of why this repo previously was the source of truth and the drift that caused: see issue Molecule-AI/molecule-core#2103.