regenerate

Pass

Audited by Gen Agent Trust Hub on Jun 27, 2026

Risk Level: SAFENO_CODE
Full Analysis
  • [SAFE]: No security issues or malicious patterns were detected. The skill is restricted to high-level instructional guidance.
  • [PROMPT_INJECTION]: The skill describes a process for ingesting external data (logs, tests, documentation), which represents a surface for indirect prompt injection. This is mitigated by the skill's own instructions to establish boundaries and provenance.
  • Ingestion points: External data sources like logs, documentation, and existing code mentioned in SKILL.md.
  • Boundary markers: Instructions include explicit steps to restate 'no-touch boundaries' and success conditions before processing data.
  • Capability inventory: No scripts or automated tools are provided within the skill; it relies on the agent's standard toolset.
  • Sanitization: The skill emphasizes human-in-the-loop verification and explicit provenance marking rather than automated filtering of external content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 27, 2026, 07:47 PM
Security Audit — agent-trust-hub — regenerate