harness-engineering

Pass

Audited by Gen Agent Trust Hub on Jun 30, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill is entirely composed of documentation and design patterns for building autonomous loops. It does not include any scripts, tools, or shell commands.
  • [SAFE]: The content specifically addresses security and safety concerns by recommending 'Locked' surfaces for evaluation and requiring explicit human approval for destructive operations or merging code.
  • [SAFE]: All external references (such as Karpathy's autoresearch or Prime Intellect) are provided as conceptual links to industry research and do not involve remote code execution or automated downloads.
  • [SAFE]: The skill includes defensive instructions against 'Metric Gaming' and indirect prompt injection by advising on the use of deterministic structure checks and source traceability.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 30, 2026, 02:29 PM
Security Audit — agent-trust-hub — harness-engineering