The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes arbitrary SKILL.md files provided as input (Phase 1). This ingestion of untrusted data represents an attack surface for instructions embedded within target files. However, the skill implements a comprehensive verification layer in Phase 3, which re-extracts functional inventory (agent calls, data structures, routing logic) from the optimized output and compares it against a pre-optimization baseline to detect and revert any unauthorized functional changes or logic loss.- [COMMAND_EXECUTION]: The skill utilizes environment-specific CLI tools including 'maestro' (for loading coding specifications and searching design patterns) and 'ccw' (for loading validation specs). These commands are used to provide context for the optimization process and are consistent with standard development environment automation.

skill-simplify