debug-with-evidence

Pass

Audited by Gen Agent Trust Hub on Sep 3, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data such as logs, traces, and source code to diagnose failures. If this external data contains malicious instructions, it could theoretically influence the agent's behavior during the 'fix' phase.
  • Ingestion points: The agent is instructed to inspect 'existing evidence', 'logs', 'traces', and 'reproduction' outputs in SKILL.md.
  • Boundary markers: The skill mandates a 'Red-capable loop' which forces the agent to verify symptoms independently before taking action, reducing the risk of blind obedience to data-embedded instructions.
  • Capability inventory: The agent has the capability to read files/logs and, when authorized, write 'smallest in-scope fixes' to the source code.
  • Sanitization: references/instrumentation.md explicitly requires that the agent 'do not capture secrets, credentials, session material, or unnecessary personal data' and defines redaction rules for production observation.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 3, 2026, 05:54 AM
Security Audit — agent-trust-hub — debug-with-evidence