The Agent Skills Directory

[SAFE]: The skill consists exclusively of documentation, templates, and evaluation cases in Markdown and JSON formats. There are no executable scripts, binary files, or configuration files that trigger external actions.
[NO_CODE]: No programming logic or package dependencies are provided. The skill relies entirely on the agent's natural language reasoning to follow the provided planning framework.
[INDIRECT_PROMPT_INJECTION]: The skill instructs agents to ingest project files (Ingestion points: project source, requirements in SKILL.md), but lacks specific boundary markers or sanitization logic. Its primary capability is planning (Capability inventory: file read and plan generation), and it explicitly advises agents to respect authorized scopes and verify surfaces, which mitigates risk.

work-decomposition