post-incident-review

Pass

Audited by Gen Agent Trust Hub on Jul 20, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: No security issues detected. The skill implements defense-in-depth measures, such as defining a restricted set of allowed tools (Read, Grep, Glob) and including explicit instructions to the agent to avoid following instructions embedded in the data being analyzed.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes external data (logs, reports), which is a potential surface for indirect prompt injection.
  • Ingestion points: The skill reads external files provided in the arguments.
  • Boundary markers: The 'Prompt Injection Safety Notice' in Section 8 explicitly tells the agent to treat analyzed directives as data rather than instructions.
  • Capability inventory: Restricted to Read, Grep, and Glob. No command execution or network capabilities.
  • Sanitization: Instructions are provided to avoid exfiltrating PII, credentials, or private keys.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 20, 2026, 02:11 PM
Security Audit — agent-trust-hub — post-incident-review