Production-grade multi-agent workflow orchestrator built with LangGraph, MCP (Model Context Protocol), A2A protocol, and PostgreSQL+pgvector. Features supervisor hub-and-spoke routing, human-in-the-loop approvals, semantic memory, circuit breakers, LLM-as-judge evaluation, and a real-time Streamlit observability dashboard.
- capability exposure inferred + 35
- recent drift inferred + 12
- tool safety inferred + 22
inferred
The A–E grade is our heuristic synthesis — a "review this" prompt, not a verdict. Each factor is tagged by what backs it: attested (a verifiable record), reported (a third party's claim), or inferred (our own heuristic, e.g. permissions). See methodology.
graded 5m ago · see ecosystem CVEs →
- C · 35 → D · 69
No known CVEs for this server.
- high dangerous code
committed secret: Slack token
- medium tool shadowing send_email
tool "send_email" shadows a verified server's tool
- medium toxic flow (lethal trifecta)
lethal trifecta reachable across this server's tools: private-data access + untrusted-content ingestion + network exfil
analyzed commit b1dd115 · analyzer v28 · 3d ago
danger signals1
- committed secret Slack token JoelJohnsonThomas-ForgeFlow-b1dd115/tests/integration/test_slack.py :40
xoxp-u…(15 chars, redacted)
- recent drift +12 capability drift →
[](https://mcpobservatory.com/servers/github:JoelJohnsonThomas/ForgeFlow/security) Heuristic, inferred signals — false positives (legitimately powerful tools, forks, language ports) are expected. Treat each as "review this", not a verdict. See the ecosystem-wide picture on the security hub, or the fleet security of JoelJohnsonThomas.