The Agent Skills Directory

[SAFE]: No malicious instructions, obfuscation techniques, or data exfiltration attempts were detected. The skill operates within the expected local environment boundaries for skill management.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard shell commands such as mkdir -p and LS to manage the .skills/ directory structure. These actions are used for local workspace organization and do not involve privilege escalation or risky parameter passing.
[SAFE]: An indirect prompt injection surface exists as the skill reads and processes external SKILL.md files for optimization and evaluation. Ingestion points: Target SKILL.md files read via the Read tool. Boundary markers: None explicitly defined. Capability inventory: mkdir, Write, Read, SearchReplace, and LS. Sanitization: None identified. This surface is considered safe given the skill's context as a structural auditing tool that does not execute the contents of the files it reads.

skill-laws