{"skill":{"slug":"smartness-eval-open-source","displayName":"Smartness Eval Open Source","summary":"OpenClaw 智能度综合评伌技能。围绕 14 个维度（含规划能力、幻觉控制）输出综合评分、证据、风险与趋势。对齐 CLEAR/T-Eval/Anthropic 行业标准。","tags":{"latest":"0.3.3"},"stats":{"comments":0,"downloads":141,"installsAllTime":0,"installsCurrent":0,"stars":0,"versions":1},"createdAt":1774165800590,"updatedAt":1774166508721},"latestVersion":{"version":"0.3.3","createdAt":1774165800590,"changelog":"- Added detailed SKILL.md including full usage guide, evaluation modes, output fields, data sources, and security declarations.\n- Documented all read/write file operations and command execution safeguards.\n- Outlined supported evaluation modes (quick, standard, deep) and their purposes.\n- Clarified new output fields (dimension_spread, trend_vs_last, pass_at_k, llm_judge).\n- Specified all data input sources and output file locations.\n- Added a Security Declaration section detailing only-read/only-write directories, limited command execution, and API key requirements for LLM Judge mode.","license":"MIT-0"},"metadata":null,"owner":{"handle":"yh22e","userId":"publishers:yh22e","displayName":"圆规","image":"https://avatars.githubusercontent.com/u/13883859?v=4"},"moderation":null}