The Agent Skills Directory

[PROMPT_INJECTION]: The skill operates by having the agent read and analyze external skill files (e.g., using cat path/to/skill/SKILL.md), which creates an indirect prompt injection attack surface (Category 8).
Ingestion points: The cat command is specified in the workflow to load external skill content into the agent's context for review.
Boundary markers: No specific delimiters or instructions are used to signal the agent to ignore or isolate instructions found within the audited files.
Capability inventory: The skill uses basic shell utilities (ls, cat, grep, wc) to facilitate the auditing process.
Sanitization: Content from the files being reviewed is not sanitized or escaped before the agent processes it for quality assessment.
[COMMAND_EXECUTION]: The skill instructions direct the agent to use standard local shell commands (ls, cat, grep, wc) to inspect directory structures and document content. These operations are limited to the local filesystem and are appropriate for the tool's auditing purpose.

skill-reviewer