Skill audit report
A Comprehensive Benchmark to Evaluate LLMs as Agents (ICLR'24)
OpenAgentSkill Trust Score
The Trust Score helps an agent decide whether a skill is safe enough to shortlist before installation.
GitHub adoption
PASS86
3.6K GitHub stars
Stars/forks activity
PASS83
3.6K stars, 270 forks; issue activity unavailable in current metadata
Recent maintenance
INFO62
7mo since push
License clarity
PASS86
Apache-2.0
README/SKILL.md completeness
INFO74
Public metadata needs stronger README/SKILL.md context
Dependency/runtime risk
PASS90
no major dependency risk hints in public metadata
Install availability
PASS92
npx skills add THUDM/AgentBench
Install command safety
PASS92
standard package or runtime install path
Permission surface
PASS100
no high-risk permission surface in public metadata
Repository evidence
PASS86
https://github.com/THUDM/AgentBench
Review status
PASS88
AI review data available
Agent Proven outcomes
INFO54
No agent outcome data yet
Checks
Install path
92
npx skills add THUDM/AgentBench
Repository
88
https://github.com/THUDM/AgentBench
License
86
Apache-2.0
Maintenance
62
7mo since push
AI review
88
Approved with no listed issues
README/SKILL.md completeness
84
Usable description available
Dependency risk
90
no major dependency risk hints in public metadata
Install command safety
92
standard package or runtime install path
Permission surface
100
no high-risk permission surface in public metadata
Stars/forks activity
83
3.6K stars, 270 forks; issue activity unavailable in current metadata
Adoption
88
3.6K GitHub stars
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.