The Agent Skills Directory

[PROMPT_INJECTION]: The skill analyzes implementation code and prototypes which serve as a potential surface for indirect prompt injection.
Ingestion points: The skill uses Read, Glob, and Grep tools to ingest data from user-defined file paths in SKILL.md (Phase 2).
Boundary markers: The skill does not define specific delimiters for separating code content from instructions, but the workflow requires a mandatory questioning phase (Phase 3) and a findings presentation phase (Phase 4) to verify intent.
Capability inventory: The skill possesses Write, Edit, and Bash capabilities across all scripts.
Sanitization: No programmatic sanitization is defined for external code content, but the "Collaborative Protocol" ensures that a human must review and approve the generated draft before the Write tool is invoked (Phase 6).
[COMMAND_EXECUTION]: The skill requests access to the Bash tool.
Evidence: allowed-tools: Bash is declared in the frontmatter of SKILL.md.
Context: The tool is utilized for analysis and project exploration. The skill's instructions guide the agent toward collaborative documentation rather than autonomous system modification.

reverse-document