The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external instruction files, which presents an indirect prompt injection surface.
Ingestion points: The skill reads SKILL.md and reference markdown files from across the skills/ directory using Glob and Read tools.
Boundary markers: There are no instructions to use specific delimiters or to wrap the content of the audited skills in a way that prevents the LLM from following instructions embedded within those files.
Capability inventory: The skill has access to Bash (for executing scripts/skill-health.sh), Write (for generating reports and JSON sidecars), and Read/Glob/Grep tools.
Sanitization: Although the skill itself checks for certain security-sensitive characters (<, >) in the frontmatter of audited skills, it does not explicitly sanitize the markdown bodies of those files before they are processed by the agent.
[COMMAND_EXECUTION]: The skill performs local command execution to gather telemetry.
Pattern: scripts/skill-health.sh report --json
Context: The skill executes a local script path to obtain usage statistics (invocations, success rates) for the skills it is auditing. This is a functional requirement of the skill.

skill-review