跳到正文
TakoAPI
返回技能列表

skill-factory

由 jeremysommerfeld8910-cpu 提供

Coding Agents & IDEs
358下载量由 clawskills.sh 提供
1浏览量
0 个赞登录后即可为这个技能点赞
—暂无评分
你的评分

简介

Create, evaluate, improve, benchmark, and publish OpenClaw skills.

描述

该技能的功能

Supports the full lifecycle of OpenClaw skill development across six modes: Create, Eval, Improve, Benchmark, Analyze, and Synthesize. Each mode targets a specific phase, from scaffolding a new skill directory to running automated quality checks and publishing to the community registry.

Replaces ad-hoc skill editing with a structured loop that ties measurable quality gates directly to the publish step, enforcing a minimum 0.80 pass rate before release.

何时使用

  • Scaffolding a new skill folder with templates and metadata
  • Running automated evals to measure a skill's assertion pass rate
  • Comparing two skill versions to identify regressions and pick a winner
  • Scanning installed skills to extract reusable trigger phrases and tool patterns
  • Publishing a finished skill to the clawhub.com community registry

示例工作流

输入

User asks: Create a skill that sends a Slack message

智能体
  1. 1

    Runs analyze_patterns.py to check for existing relevant patterns in installed skills

  2. 2

    Writes a one-paragraph spec covering trigger conditions, happy path, and output format

  3. 3

    Runs init_skill.py to scaffold the skill folder with SKILL.md and subdirectories

  4. 4

    Writes SKILL.md frontmatter and body following progressive disclosure rules

  5. 5

    Packages the skill with package_skill.py and runs the eval suite to verify pass rate

输出

A packaged slack-notifier.skill file with passing evals and a run report showing pass_rate >= 0.80

前置要求

  • clawhub.com account (required for publishing skills)

安装

clawhub install jeremysommerfeld8910-cpu/skill-factory

运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。

链接