The Agent Skills Directory

[SAFE]: The skill serves as a best-practice baseline for Python development. It recommends using modern, reputable tools such as uv for dependency management and ruff for linting and formatting.
[INDIRECT_PROMPT_INJECTION]: In references/data.md, the skill identifies untrusted input (e.g., scraped content, webhooks) as an injection surface. It provides defensive guidance, instructing developers to explicitly treat such input as untrusted and to avoid instructions that a model might follow if they were embedded in the data.
Ingestion points: External data sources like scraped content and webhooks mentioned in references/data.md.
Boundary markers: Recommends explicit code or model instructions to treat inputs as untrusted.
Capability inventory: No dangerous execution capabilities provided; guidelines focus on safe SDK wrappers.
Sanitization: Recommends Pydantic validation and schema checks at the ingestion boundary.
[COMMAND_EXECUTION]: The skill provides several standard command-line examples for project initialization, dependency management, and testing (e.g., uv sync, uv run ruff check, uv run pytest). These are strictly for developer workflow and do not involve unsafe handling of user input.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known Python libraries (e.g., FastAPI, Polars, Pydantic) and uses the uv package manager, which is a standard tool in the Python ecosystem. No suspicious third-party scripts or untrusted repositories are referenced.

z-py-core