peer-review

Pass

Audited by Gen Agent Trust Hub on Aug 4, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data (manuscripts and supplementary materials). This creates a surface for indirect prompt injection where instructions embedded in a manuscript could attempt to influence agent behavior.
  • Ingestion points: Manuscript files or pasted text identified in Step 1 of the workflow.
  • Boundary markers: The instructions do not explicitly mandate the use of delimiters or 'ignore instructions' wrappers when processing the manuscript text.
  • Capability inventory: The skill has access to tools including Bash, Write, and Edit, which represent a significant capability tier if an injection were successful.
  • Sanitization: No explicit content sanitization steps are defined for the manuscript text.
  • [CREDENTIALS_UNSAFE]: The skill includes proactive security instructions for 'Enhanced Advisory Mode', mandating that API keys must be kept in environment variables and never written to review files, aligning with secret management best practices.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 4, 2026, 09:25 PM
Security Audit — agent-trust-hub — peer-review