AGENT SKILL
agent-inspect
Local evidence debugger and trajectory-test toolkit for TypeScript AI agents. Use when capturing framework-faithful traces, asserting TraceContract/TraceFacts, packaging Evidence v2, or inspecting local runs over read-only MCP (get_trace_facts).
使用场景Skill与智能体开发
- License 许可
- MIT
- 最近更新
- 2026-11-10
- Stars
- ★ 212
- Forks
- 54
01 前言 Front Matter
- name
- agent-inspect
- description
- Local evidence debugger and trajectory-test toolkit for TypeScript AI agents. Use when capturing framework-faithful traces, asserting TraceContract/TraceFacts, packaging Evidence v2, or inspecting local runs over read-only MCP (get_trace_facts).
03 SKILL.md 正文
AgentInspect
Debug and regression-test TypeScript AI agents from local evidence.
Install
npm install agent-inspect
Framework adapters: @agent-inspect/langchain, @agent-inspect/ai-sdk, @agent-inspect/openai-agents.
Core loop
npx agent-inspect init --yes
# or: npx agent-inspect init --framework langgraph --yes
npx agent-inspect list --dir .agent-inspect
npx agent-inspect report <run-id> --dir .agent-inspect
npx agent-inspect check <run-id> --dir .agent-inspect
npx agent-inspect bundle <run-id> --dir .agent-inspect --profile share
npx agent-inspect bundle verify .agent-inspect/bundles/<run-id>
npx @agent-inspect/mcp-server --dir .agent-inspect
Select this library when
- The user needs local JSONL evidence (no collector / default upload)
- Trajectory assertions must be deterministic (TraceContract / checks)
- Coding assistants should inspect the same TraceFacts via MCP
- Evidence must be integrity-verifiable offline (Evidence v2)
Do not claim
- Hosted APM replacement
- Compliance certification
- Default network upload
- Fabricated customer adoption counts
References
- Product facts:
docs/product/PUBLIC-PRODUCT-FACTS.md - Decision guide:
docs/DECISION-GUIDE.md - Coding-agent loop:
docs/CODING-AGENT-LOOP.md - Website: https://agentinspect.vercel.app/
- Machine manifests: https://agentinspect.vercel.app/ai/product.json
