global-reviewer

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because it ingests and evaluates untrusted data from survey drafts.
  • Ingestion points: The skill reads output/DRAFT.md and auxiliary files in the outline/ directory (e.g., outline.yml, taxonomy.yml) to perform its review.
  • Boundary markers: No explicit boundary markers or instructions to disregard embedded commands are utilized when processing the draft content within the LLM role prompt context.
  • Capability inventory: The skill has the capability to write to output/GLOBAL_REVIEW.md and apply modifications to output/DRAFT.md.
  • Sanitization: There is no evidence of input sanitization or filtering applied to the draft content before it is processed by the AI or analyzed by the script logic.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 01:59 PM