The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references the skills-ref validation tool hosted on GitHub at https://github.com/agentskills/agentskills/tree/main/skills-ref. This is a reference to a tool intended for specification validation and follows standard development practices.
[COMMAND_EXECUTION]: Instructions include running the skills-ref validate <skill-path> command. This is a legitimate use of a CLI tool for its primary purpose of validating file structures against the agentskills.io specification.
[PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data in the form of third-party skill files (SKILL.md, scripts, and instruction files like AGENTS.md). This creates a surface for indirect prompt injection where malicious instructions embedded in the audited files could attempt to influence the agent's auditing logic.
Ingestion points: Reads SKILL.md, CLAUDE.md, AGENTS.md, and any referenced files during the audit process.
Boundary markers: None explicitly defined to isolate audited content from instructions.
Capability inventory: Performs file system listing and reading, and executes the skills-ref validation tool.
Sanitization: No specific sanitization or filtering of external content is mentioned before processing.

improving-skills