Unified AI agent framework: one interface for 9 LLM providers (Qwen, Kimi, GLM, DeepSeek, MiniMax, Doubao, ChatGPT, Gemini, Claude) with tools, MCP, sub-agents, skills, RAG knowledge base, scheduler, multi-channel IM gateway (WeChat/Feishu/Telegram) and multi-user serving
- capability exposure inferred + 35
- tool safety inferred + 5
- trust mitigators mixed − 3
inferred mixed
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 42s ago · see ecosystem CVEs →
No known CVEs for this server.
- medium dangerous code
dynamic exec: eval()/exec()
analyzed v0.5.2 · analyzer v32 · 1h ago
skills & prompt files 10
- agent-rules milu-0.5.2/src/milu/templates/prompts/coder/agent.md
- agent-rules milu-0.5.2/src/milu/templates/prompts/main/agent.md
- agent-rules milu-0.5.2/src/milu/templates/prompts/reader/agent.md
- agent-rules milu-0.5.2/src/milu/templates/prompts/researcher/agent.md
- agent-rules milu-0.5.2/src/milu/templates/prompts/reviewer/agent.md
- skill milu-0.5.2/src/milu/templates/skills/frontend-design/SKILL.md
- skill milu-0.5.2/src/milu/templates/skills/internal-comms/SKILL.md
- skill milu-0.5.2/src/milu/templates/skills/mcp-builder/SKILL.md
- skill milu-0.5.2/src/milu/templates/skills/systematic-debugging/SKILL.md
- skill milu-0.5.2/src/milu/templates/skills/test-driven-development/SKILL.md
danger signals5
- dynamic code execution eval()/exec() milu-0.5.2/examples/2. agent_basic.py :28
return str(eval(expression)) # noqa: S307 - dynamic code execution eval()/exec() milu-0.5.2/src/milu/sandbox/local.py :63
exec(code, restricted_globals) # noqa: S102 - dynamic code execution eval()/exec() milu-0.5.2/tests/test_agent_integration.py :32
result = eval(expression, {"__builtins__": {}}, {}) - dynamic code execution eval()/exec() milu-0.5.2/tests/test_real_new_providers.py :118
return str(eval(expression)) - suspicious endpoint api.telegram.org milu-0.5.2/src/milu/channels/telegram.py :40
api_base: str = "https://api.telegram.org"
[](https://mcpobservatory.com/servers/pypi:milu/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 stephon.