The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow extensively utilizes uv run to execute local Python scripts (e.g., skf-resolve-package.py, skf-extract-public-api.py, skf-render-quick-metadata.py) that perform specialized source analysis and document assembly tasks. It also employs standard shell commands for filesystem management, including mkdir, rm, and rmdir to handle temporary run directories and versioned output packages.
[EXTERNAL_DOWNLOADS]: The skill fetches repository source files and metadata from GitHub's official raw content domain (raw.githubusercontent.com) and utilizes registries like npm, PyPI, and crates.io for package resolution. Additionally, it invokes npx skill-check to perform standardized validation via the npm registry. All these domains are recognized as trusted or well-known services.
[INDIRECT_PROMPT_INJECTION]: As the skill is designed to ingest and process third-party repository content (such as README files and manifest descriptions) to generate new skill documentation, it possesses an attack surface for indirect prompt injection. A malicious repository could include deceptive instructions in its documentation. The skill mitigates this by using automated extraction scripts, intermediate structured JSON formats, and strict templates for the final output.
Ingestion points: README.md, package manifests, and source code entry points fetched in references/quick-extract.md.
Boundary markers: Instructions strictly define the SKILL.md structure and require the use of a provided template (assets/skill-template.md).
Capability inventory: File system writes (references/write-and-validate.md), command execution via uv run/npx, and network access for data fetching and web search.
Sanitization: Extraction data is processed into structured JSON before being rendered into the final document, and instructions mandate proper JSON escaping for metadata fields.
[DYNAMIC_EXECUTION]: The skill includes an on_complete hook feature, allowing for the execution of a user-defined shell command after successful generation. This feature is intended for workflow integration (e.g., triggering a git commit or registry update) and is disabled by default.

skf-quick-skill