The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references several Python packages for the migration target, including pydantic-ai, pydantic-ai-harness, pydantic-ai-slim, and pydantic-evals. These are official resources for the Pydantic AI ecosystem.
[COMMAND_EXECUTION]: The migration instructions discuss using tools such as Shell and Coder within the Pydantic AI Harness. The skill provides defensive guidance, recommending the use of sandboxes (containers or VMs) for untrusted code execution to prevent unauthorized system access.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external repositories during the migration analysis.
Ingestion points: The agent is instructed in SKILL.md to read repository instructions, code dependencies, and runtime entrypoints from the source application.
Boundary markers: The process requires establishing characterization tests and baseline proofs for each migrated contract to ensure behavior parity.
Capability inventory: Includes tool execution, file system access, and agent loop operations provided by the Pydantic AI core.
Sanitization: The skill explicitly advises mapping sensitive context and authenticated identities to typed dependencies rather than exposing them as model-controllable tool inputs.
[DYNAMIC_EXECUTION]: The instructions describe the use of Harness Skills for loading on-demand instructions and Coder for automated refactoring. These represent standard framework capabilities for managing agent behavior and code generation during the migration workflow.

migrating-mastra-to-pydantic-ai