Capability Evaluation

在固定任务与明确标签下对照能力策略,记录正确率、漏检、误报、脚本纠正与耗时;不把微型离线实验当真实模型或工程收益。

339 stars 13 forksOther

Install

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add Lanqiao-Luca/ai-java-workflow --skill capability-evaluation
Manual install
git clone --depth 1 https://github.com/Lanqiao-Luca/ai-java-workflow.git /tmp/ai-java-workflow
mkdir -p ~/.claude/skills
cp -r /tmp/ai-java-workflow/skills/capability-evaluation ~/.claude/skills/capability-evaluation

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

**目的**:将明确评估问题适配成同输入实验与可重算证据,交付限定结果和未测维度。

**适用于**:责任 Agent 已接收任务集合、策略、判定依据、环境与运行预算的有限对照实验。

# 能力对照评估

消费[CE01–CE05](../../standards/capability-evaluation.md),先固定任务/标签及基线意义,核同输入、初始判定与纠正阶段分开。按[验证](../../standards/verification.md)记录真实执行与未测项。

对仓库首批反证报告任务,可使用[基准工具](../../tools/capability_benchmark.py)和[固定案例](../../workflows/examples/capability-benchmark/README.md)。工具只比较报告结构判定;基线是明确故意简化的字段检查,不是模型、用户原实践或外部产品。

逐项保初始判断、标签、状态、计时、预设纠正次数;复算漏检/误报及分母。脚本还原不等人工返工,单次耗时不宣传提速;没有测量的模型成本、人工改码和业务效果保 NOT_MEASURED。

输出任务/策略/输入摘要、原始观察、汇总、限制和下一接收者。候选变化重建实验;不自动采私密数据、调用模型、对外发布或后台复跑。

返回:[Skills](../README.md)。
Read the full skill on GitHub

Details

Skill name
capability-evaluation
Path
skills/capability-evaluation/SKILL.md
License
MIT
Repo updated
17 November 2026

FAQ

What does the Capability Evaluation skill do?

在固定任务与明确标签下对照能力策略,记录正确率、漏检、误报、脚本纠正与耗时;不把微型离线实验当真实模型或工程收益。

How do I install the Capability Evaluation skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add Lanqiao-Luca/ai-java-workflow --skill capability-evaluation. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Capability Evaluation popular?

It lives in Lanqiao-Luca/ai-java-workflow, which has 339 GitHub stars and 13 forks.

Related skills

Browse the most popular Claude skills, all other skills, or write your own with the skills generator.