migrating-langchain-to-pydantic-ai
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted codebase data (source code, instructions, and tests) to perform migrations, which creates a potential surface for indirect prompt injection.
- Ingestion points: The agent is instructed to read repository files and dependency definitions in SKILL.md and references/VERIFICATION-AND-CUTOVER.md.
- Capability inventory: The migration process involves using Pydantic AI tools including filesystem access, shell execution (via pydantic_ai_harness), and subagent delegation.
- Boundary markers: The documentation provides explicit guidance on establishing security boundaries, such as using typed dependencies to isolate model inputs from service clients.
- Sanitization: The instructions mandate validation of terminal choices and the use of adapters to sanitize external content.
- [EXTERNAL_DOWNLOADS]: The skill references and encourages the use of several external packages and services.
- Evidence: Mentions of pydantic-ai, pydantic-ai-harness, logfire, langchain, langgraph, and deepagents.
- Evidence: Mentions of durable execution integrations such as DBOS, Temporal, Prefect, and Restate.
- Note: These resources are either vendor-official or well-known services within the ecosystem.
Audit Metadata