The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional markdown and a reference rubric. It does not perform network requests, execute shell commands, or access sensitive local files.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface as it is designed to ingest and evaluate external files from the skill package being reviewed. 1. Ingestion points: SKILL.md, bundled references, and scripts of the target skill. 2. Boundary markers: Absent. 3. Capability inventory: Limited to analytical evaluation; no subprocess calls, network operations, or file-write permissions were detected across the skill files. 4. Sanitization: Absent. The risk is considered safe as the instructions direct the agent to follow a structured, evidence-based auditing process which helps maintain objectivity despite processing untrusted content.

skill-judge