The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were identified in the skill files. The instructions prioritize safe data handling and user confirmation.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface due to its role in processing external skill files and repository data.
Ingestion points: Reads and analyzes existing skill definitions and discovery paths in SKILL.md.
Boundary markers: Instructions explicitly require identifying conflicts and stopping for user confirmation rather than automated resolution.
Capability inventory: Includes file system operations (moving files, symlinking) and execution of local validation utilities.
Sanitization: Employs structural validation and mandates observable output evaluations to confirm the intended behavior.
[COMMAND_EXECUTION]: Instructs the agent to utilize standard environment commands (e.g., claude --version) and local validators to confirm tool availability and skill compliance, which are appropriate for its administrative purpose.

to-skill