The Agent Skills Directory

[COMMAND_EXECUTION]: The workflow performs extensive command execution using uv run to invoke specialized local Python scripts such as skf-extract-public-api.py, skf-detect-workspaces.py, and skf-write-skill-brief.py. These scripts handle the technical heavy lifting of parsing codebases and managing the file lifecycle.
[EXTERNAL_DOWNLOADS]: The skill uses gh api and curl to fetch repository metadata, file trees, and manifest contents from GitHub. These network operations are strictly limited to the target repositories specified by the user and are used solely for the scoping analysis.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from external repositories (e.g., README files and manifests) to synthesize skill descriptions. Evidence Chain: 1. Ingestion: step-02-analyze-target.md fetches repository contents. 2. Boundaries: The workflow includes mandatory user confirmation gates in step-01, step-03, and step-04. 3. Capabilities: The skill can execute local scripts and write to the filesystem. 4. Sanitization: It utilizes dedicated scripts for technical parsing rather than relying on the LLM to process raw external content directly.

skf-brief-skill