skill-review

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external content to perform a quality audit, which creates a surface for indirect prompt injection if the target files contain instructions designed to manipulate model behavior.\n
  • Ingestion points: The skill reads the target SKILL.md and other files in the skill directory as specified in SKILL.md (Step 1.2).\n
  • Boundary markers: Absent. The workflow in SKILL.md does not define delimiters to separate ingested content from skill instructions.\n
  • Capability inventory: The skill performs file reading (SKILL.md), workspace directory listing (SKILL.md), and progressively loads internal reference files like references/review-criteria.md. It lacks network capabilities or destructive file system access.\n
  • Sanitization: Absent. No validation or filtering is specified for the ingested markdown content.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 11:00 AM
Security Audit — agent-trust-hub — skill-review