Agent Evaluation
技能 已验证 活跃Use when testing skills, commands, or agents for quality. Use after creating new skills, before deploying agents, or when debugging inconsistent agent behavior. Triggers on "evaluate", "test quality", "is this skill working", or QA of AI workflows.
Systematically assess and improve the quality and reliability of AI agents and skills through a structured evaluation process.
功能
- Structured 5-dimension evaluation rubric
- Methods for direct scoring and LLM-as-judge
- Bias detection and mitigation strategies
- Workflow examples for new skills and agent QA
- Test case design principles including edge cases
使用场景
- Evaluating new skills before deployment
- Debugging inconsistent agent behavior
- Comparing different agent approaches or models
- Performing systematic QA reviews of AI workflows
非目标
- Performing one-off manual reviews of simple outputs
- Evaluating purely creative or subjective tasks
- Adding significant latency to real-time evaluations
工作流
- Define criteria and thresholds
- Create test cases (easy, medium, hard, adversarial)
- Run direct scoring or LLM-as-judge evaluation
- Compare outputs and validate against ground truth
- Monitor agreement and iterate on prompts/skills
实践
- Evaluation methodology
- Quality assurance
- Agent testing
- Rubric design
安装
npx skills add guia-matthieu/clawfu-skills通过 npx 运行 Vercel skills CLI(skills.sh)— 需要本地安装 Node.js,以及至少一个兼容 skills 的智能体(Claude Code、Cursor、Codex 等)。前提是仓库遵循 agentskills.io 格式。
质量评分
已验证类似扩展
Telegram Crabbox E2e Proof
100Use when reviewing, reproducing, or proving OpenClaw Telegram behavior with a real Telegram user on Crabbox, including PR review workflows that need an agent-controlled Telegram Desktop recording, TDLib user-driver commands, Convex-leased credentials, WebVNC observation, and motion-trimmed artifacts.
Review Skill Format
99Review a SKILL.md file for compliance with the agentskills.io standard. Checks YAML frontmatter fields, required sections, line count limits, procedure step format, and registry synchronization. Use when a new skill needs format validation before merge, an existing skill has been modified and requires re-validation, performing a batch audit of all skills in a domain, or reviewing a contributor's skill submission in a pull request.
Init
100创建或优化存储库的 AGENTS.md 文件，提供最少、高信号的说明，涵盖代理无法从代码库推断的不可发现的编码约定、工具怪癖、工作流偏好和项目特定规则。在为新存储库设置代理说明或 Claude 配置时，当现有的 AGENTS.md 文件过长、通用或过时，当代理反复犯可避免的错误，或当存储库工作流发生变化且需要修剪代理配置时使用。应用可发现性过滤器—省略 Claude 可从 README、代码、配置或目录结构中学到的任何内容—并应用质量门，以验证每行是否仍然准确且具有操作意义。
Context Compression
100This skill should be used when the user asks to "compress context", "summarize conversation history", "implement compaction", "reduce token usage", or mentions context compression, structured summarization, tokens-per-task optimization, or long-running agent sessions exceeding context limits.
Qa
100Interactive QA session where user reports bugs or issues conversationally, and the agent files GitHub issues. Explores the codebase in the background for context and domain language. Use when user wants to report bugs, do QA, file issues conversationally, or mentions "QA session".
Context7 Cli
100使用 ctx7 CLI 获取库文档、管理 AI 编码技能并配置 Context7 MCP。当用户提及“ctx7”或“context7”时，需要任何库的当前文档、希望安装/搜索/生成技能，或需要为 AI 编码代理设置 Context7 时激活。