The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run project-local validation tools like pnpm lint and skills-ref validate to ensure skill formatting adheres to specifications.
[EXTERNAL_DOWNLOADS]: The skill performs network requests via web search or document fetching to verify factual claims, such as API versions and command flags, against official sources.
[PROMPT_INJECTION]: The skill audits content from external directories, creating an attack surface for instructions embedded in audited files (indirect prompt injection).
Ingestion points: The skill reads all content within SKILL.md and reference files from target directories like .claude/skills/ or user-specified paths.
Boundary markers: No explicit delimiters or instructions are used to isolate untrusted skill content from the auditing agent's logic.
Capability inventory: The skill has the capability to execute CLI commands (pnpm, skills-ref), perform network searches, and write updates to local skill files.
Sanitization: No explicit sanitization or filtering of the audited content is mentioned.

skill-auditing