engineer-system-change

Pass

Audited by Gen Agent Trust Hub on Aug 15, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides high-level engineering logic and lacks any executable scripts, remote downloads, or credential-accessing commands.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and act upon external data such as RFCs and codebases.\n
  • Ingestion points: Processes external design documents and source code mentioned in SKILL.md.\n
  • Boundary markers: Instructions mandate the separation of verified facts from inferences and the use of explicit decision gates.\n
  • Capability inventory: The skill assumes access to tools for reading and implementing code changes.\n
  • Sanitization: The framework requires executable verification plans, including adversarial checks and regression tests, for all changes.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 15, 2026, 02:59 AM
Security Audit — agent-trust-hub — engineer-system-change