The Agent Skills Directory

[COMMAND_EXECUTION]: Evaluated the use of subprocess.run within the testing guidelines in references/modules.md. The script provides a template for testing module import integrity by launching a fresh interpreter to verify each package member. This is a standard software engineering practice for detecting circular dependencies and import-time side effects, and it does not allow for arbitrary command injection or unauthorized execution.
[INDIRECT_PROMPT_INJECTION]: Analyzed the data ingestion and parsing strategies outlined in references/parsing.md. The skill promotes a 'parse, don't validate' approach using Pydantic, specifically recommending strict=True and extra="forbid" configurations to prevent type coercion and unexpected payload data. These instructions serve to harden applications against untrusted data rather than creating a vulnerability surface.
[EXTERNAL_DOWNLOADS]: The skill references several well-known and trusted Python libraries in references/toolchain.md (e.g., pydantic, msgspec, stamina, anyio). These are standard ecosystem dependencies, and the skill provides specific version pinning advice to avoid known compatibility issues, particularly regarding Python 3.14, which is a positive security practice.
[CREDENTIALS_UNSAFE]: Analyzed the configuration management advice in references/ports-persistence.md. The skill correctly identifies the risks of ambient environment access and recommends centralizing all configuration parsing into a single 'composition root' using Pydantic's SecretStr to prevent accidental credential leakage in logs or tracebacks.
[PROMPT_INJECTION]: The instructions contain no attempts to override agent constraints or bypass safety filters. The language is purely technical and focused on software design principles.

python