learn-from-math-agent-trajectories

Pass

Audited by Gen Agent Trust Hub on Oct 26, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data sources, including mathematical agent trajectories, transcripts, and scratch code. These sources represent an attack surface where malicious instructions could be embedded to influence the agent's behavior during the review process.
  • Ingestion points: Trajectories, transcripts, scratch code, and artifacts from previous agent sessions.
  • Boundary markers: The instructions do not specify explicit delimiters for the trajectory data, though they emphasize an analytical posture.
  • Capability inventory: The skill identifies routing to other tools (e.g., evaluate-mcp-tool-adoption) and suggests updates to skills and documentation.
  • Sanitization: No explicit sanitization or filtering of the processed trajectory content is mentioned.
  • [SAFE]: The skill includes a clear safety guideline requiring explicit user authorization before performing any external mutations or repository changes suggested by the analysis.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 26, 2026, 03:38 PM