Skill audit report
Official AHE code — Agentic Harness Engineering: observability-driven automatic evolution of coding-agent harnesses (concurrent w/ meta-harness). NexAU-AHE reaches 84.7% ± 2.1 pass@1 on Terminal-Bench 2 (GPT-5.5). Lifts GPT-5.4 69.7→77.0% over 10 iters, beats Codex/ACE/Training-Free GRPO; frozen harness transfers to SWE-bench-Verified.
OpenAgentSkill Trust Score
The Trust Score helps an agent decide whether a skill is safe enough to shortlist before installation.
GitHub adoption
INFO76
600 GitHub stars
Stars/forks activity
INFO71
600 stars, 67 forks; issue activity unavailable in current metadata
Recent maintenance
INFO76
3mo since push
License clarity
PASS86
MIT
README/SKILL.md completeness
PASS90
Metadata includes enough usage and workflow context
Dependency/runtime risk
INFO72
command execution surface
Install availability
PASS92
npx skills add china-qijizhifeng/agentic-harness-engineering
Install command safety
PASS92
standard package or runtime install path
Permission surface
INFO62
shell or command execution, filesystem or document access
Repository evidence
PASS86
https://github.com/china-qijizhifeng/agentic-harness-engineering
Review status
PASS88
AI review data available
Agent Proven outcomes
INFO54
No agent outcome data yet
Checks
Install path
92
npx skills add china-qijizhifeng/agentic-harness-engineering
Repository
88
https://github.com/china-qijizhifeng/agentic-harness-engineering
License
86
MIT
Maintenance
76
3mo since push
AI review
88
Approved with no listed issues
README/SKILL.md completeness
90
Usable description available
Dependency risk
72
command execution surface
Install command safety
92
standard package or runtime install path
Permission surface
62
shell or command execution, filesystem or document access
Stars/forks activity
71
600 stars, 67 forks; issue activity unavailable in current metadata
Adoption
88
600 GitHub stars
Financial decision safety
58
Research-only use: do not treat output as financial advice or execute a position without human approval.
Warnings
Method
This report combines public metadata, AI review output, repository freshness, install readiness, OpenAgentSkill events, quality scoring, trust checks, and the agent safety gate. It is not a full source-code security review.