The Agent Skills Directory

[SAFE]: The skill provides standard documentation and templates for Python package development. It includes appropriate disclaimers and security advice regarding the management of API tokens and distribution artifacts. The use of placeholders like 'YOUR_API_KEY_HERE' and the recommendation to use environment secrets for GitHub Actions are examples of secure instruction design.
[INDIRECT_PROMPT_INJECTION]: The skill contains a vulnerability surface where user-provided metadata (e.g., package name, description) is used to generate files and shell commands. This is an inherent property of developer-oriented skills and is mitigated by instructions to validate outcomes and seek clarification on safety boundaries.
Ingestion points: Project metadata (names, descriptions, version strings) provided by the user and processed via the templates in SKILL.md and resources/implementation-playbook.md.
Boundary markers: Present. SKILL.md contains a 'Limitations' section explicitly directing the agent to stop and ask for clarification if safety boundaries are missing and to avoid treating output as a substitute for expert review.
Capability inventory: File system operations for creating directory structures and writing configuration files; execution of build and distribution tools (pip, python -m build, twine) as documented in the implementation playbook.
Sanitization: The skill recommends the 'src' layout for better project isolation and relies on modern packaging standards (PEP 517/518/621) which use structured metadata formats like TOML.

python-packaging