The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to utilize repository-specific tooling for validation and management, including executing basedpyright (via binary or uvx), and running just commands such as just check, just bundle, and just typecheck.
[DYNAMIC_EXECUTION]: Guidelines specify the use of importlib.import_module combined with getattr to load build-only modules that are excluded from wheels, primarily to resolve type-checking diagnostics for modules the checker cannot statically resolve.
[INDIRECT_PROMPT_INJECTION]: As a tool for processing code changes and user requests, the skill operates on an untrusted data surface. It proactively mitigates this risk by instructing the agent to validate untrusted input at the boundary, requiring expected container shapes, and converting validated data into typed trusted representations (frozen dataclasses or domain value types) before further processing.
[SAFE]: The skill promotes several defensive coding practices:
Supply Chain Security: Enforces a 'stdlib-first' policy and requires all third-party dependencies to be pure Python and explicitly audited/added to a specific requirements file.
Input Validation: Mandates strict validation and parsing of external data at the system boundary to prevent downstream injection or logic errors.
Environment Hermeticity: Requires tests to be self-contained within temporary directories, specifically prohibiting dependencies on network access or external state.

python-authoring