agentenatalie/skill-bridge

最近记录的来源活动: 2026年3月31日 01:02

SkillsMP 收录数据更新: 2026年11月5日 19:57

已收集 skills: 4

GitHub 星标: 0
GitHub Forks: 0

用 AI 助手安装

复制这段 Prompt，发给你正在使用的 AI 助手。

Follow https://skillsmp.com/skill-install/prompt.md to install Agent Skills from https://github.com/agentenatalie/skill-bridge.

仓库 README

仓库 README 来源：README.md

查看 README 源文件

展开完整 README

<div align="center"> # SkillBridge [![CI](https://img.shields.io/github/actions/workflow/status/agentenatalie/skill-bridge/ci.yml?branch=main&label=CI)](https://github.com/agentenatalie/skill-bridge/actions/workflows/ci.yml) [![Release](https://img.shields.io/github/actions/workflow/status/agentenatalie/skill-bridge/release.yml?label=Release)](https://github.com/agentenatalie/skill-bridge/actions/workflows/release.yml) [![License](https://img.shields.io/badge/license-MIT-black)](./LICENSE) [![Node](https://img.shields.io/badge/node-%3E%3D20-339933)](https://nodejs.org/) [![中文文档](https://img.shields.io/badge/README-%E4%B8%AD%E6%96%87-1677ff)](./README.zh-CN.md) </div> SkillBridge is a tool for moving skills between agent ecosystems. If you already have a good skill, you should not have to rewrite it from scratch just to use it in another tool. SkillBridge reads the source skill, converts what it can, tells you what does not map cleanly, and writes the target format back out. In plain English: > Take one skill and migrate it to another ecosystem. It stays focused on one job: skill migration. ## Why this exists The same idea now shows up in a lot of different skill formats: - OpenClaw `SKILL.md` - Claude Code skills - Codex skills with optional `agents/openai.yaml` - OpenCode skills - `AGENTS.md` - Cursor rules - Cline rules They all look similar at a glance, but they differ in file layout, frontmatter, tool names, sidecars, and ecosystem-specific rules. Doing that translation by hand is boring. Doing it repeatedly is error-prone. SkillBridge exists to make that step fast and predictable. ## What you get - CLI-first workflow - scored format detection - `analyze`, `convert`, `fix`, `formats`, and `corpus` commands - deterministic parser / converter / analyzer core - optional Codex sidecar preservation and synthesis - bundle output for targets that need companion files - local corpus regression checks - a small web playground that uses the same engine ## Supported formats | Format | Support | Notes | | --- | --- | --- | | `openclaw` | yes | deterministic compatibility profile | | `claude` | yes | Claude Code skill profile | | `codex` | yes | official-style profile with optional `agents/openai.yaml` | | `opencode` | yes | OpenCode skill profile | | `agentsmd` | yes | `AGENTS.md` adapter | | `cursor` | yes | Cursor rules adapter | | `cline` | yes | Cline rules adapter | ## Quick examples Detect a skill format: ```bash node dist/src/cli.js detect examples/codex/figma-sync/SKILL.md ``` Analyze an OpenClaw skill before converting it to Claude Code: ```bash node dist/src/cli.js analyze \ examples/openclaw/SKILL.md \ --from openclaw \ --target claude ``` Convert a Claude skill into OpenCode: ```bash node dist/src/cli.js convert \ examples/claude/git-release/SKILL.md \ --from claude \ --to opencode ``` Convert an OpenCode skill into a Codex bundle: ```bash node dist/src/cli.js convert \ examples/opencode/figma-bridge/SKILL.md \ --from opencode \ --to codex \ --bundle-dir ./out/figma-bridge ``` That writes: - `./out/figma-bridge/SKILL.md` - `./out/figma-bridge/agents/openai.yaml` ## Install Node `20+` is required. ```bash npm install npm run build ``` If you are working from the repo, use: ```bash node dist/src/cli.js <command> ``` If you install the package globally later, the command is: ```bash skillbridge <command> ``` ## Commands ```bash skillbridge detect <path> [--from auto|<format>] [--json] skillbridge analyze <path> --from <format|auto> --target <format> [--json] skillbridge convert <path> --from <format|auto> --to <format> [-o output.md] [--bundle-dir dir] skillbridge fix <path> --from <format|auto> --target <format> [-o output.md] [--bundle-dir dir] skillbridge corpus [--limit 20] [--json] skillbridge formats ``` What they do: - `detect`: guess the source format and show scored candidates - `analyze`: report compatibility issues before conversion - `convert`: generate target output using deterministic mappings - `fix`: same as `convert`, but adds compatibility notes for unresolved parts - `corpus`: sample local skills and make sure detection / analysis / conversion still work - `formats`: print supported formats, expected paths, frontmatter, and sidecars ## How it works SkillBridge is intentionally simple inside: 1. Parse the source skill 2. Normalize it into an internal spec 3. Convert that spec into the target format 4. Analyze what mapped cleanly and what did not The normalized spec currently tracks: - `name` - `description` - `instructions` - `tools` - `inputs` - `constraints` - `examples` - sidecars ## Deterministic by default The important part of SkillBridge is rule-based. - parsing is deterministic - conversion is deterministic - compatibility reporting is deterministic - unsupported fields are surfaced instead of silently guessed That is the whole point. This should feel like a migration tool, not a black box rewrite bot. ## Codex sidecar support SkillBridge supports the optional Codex sidecar: - `agents/openai.yaml` Current behavior: - preserve an existing sidecar when the source already has one - synthesize one when the source contains enough structured metadata - write it out with `--bundle-dir` Currently synthesized fields: - `interface.display_name` - `interface.short_description` - `interface.default_prompt` - `interface.icon_small` - `interface.icon_large` - `dependencies.tools[*].type` - `dependencies.tools[*].value` - `dependencies.tools[*].description` - `dependencies.tools[*].transport` - `dependencies.tools[*].url` ## Web playground There is also a small playground in [`playground/`](./playground/). Run it locally: ```bash npm run build npm run playground:serve ``` Then open: - `http://127.0.0.1:4173/playground/index.html` It uses the same browser-side engine as the CLI and gives you: - source editor - output viewer - raw / view / diff modes - detection / issues / sidecars panels - English and Simplified Chinese UI ## Examples in this repo - [`examples/openclaw/SKILL.md`](./examples/openclaw/SKILL.md) - [`examples/claude/git-release/SKILL.md`](./examples/claude/git-release/SKILL.md) - [`examples/codex/figma-sync/SKILL.md`](./examples/codex/figma-sync/SKILL.md) - [`examples/codex/figma-sync/agents/openai.yaml`](./examples/codex/figma-sync/agents/openai.yaml) - [`examples/opencode/git-release/SKILL.md`](./examples/opencode/git-release/SKILL.md) - [`examples/opencode/figma-bridge/SKILL.md`](./examples/opencode/figma-bridge/SKILL.md) - [`examples/agentsmd/AGENTS.md`](./examples/agentsmd/AGENTS.md) - [`examples/cursor/.cursor/rules/release-policy.mdc`](./examples/cursor/.cursor/rules/release-policy.mdc) - [`examples/cline/.clinerules/research.md`](./examples/cline/.clinerules/research.md) ## Verify before publishing ```bash npm run build npm test npm run test:corpus npm pack --dry-run ``` ## Repo layout ```text src/ CLI and conversion engine examples/ sample skills playground/ static web UI test/ regression tests ``` ## CI and release This repo already includes: - [`ci.yml`](./.github/workflows/ci.yml) for build and test - [`release.yml`](./.github/workflows/release.yml) for npm publish - [`pages.yml`](./.github/workflows/pages.yml) for GitHub Pages `prepublishOnly` runs `release:check`, which runs tests and `npm pack --dry-run`. ## References The current adapters are grounded in: - [OpenAI Codex skills docs](https://developers.openai.com/codex/skills) - [OpenAI Codex rules docs](https://developers.openai.com/codex/rules) - [OpenCode skills docs](https://opencode.ai/docs/skills) - [OpenCode config docs](https://opencode.ai/docs/bs/config/) ## Roadmap - broader sidecar support beyond Codex `agents/openai.yaml` - more adapter-specific normalization rules - optional AI-assisted rewrite mode outside the deterministic core - richer playground comparison modes ## License MIT

正在加载仓库 README…

这个仓库中的 skills

软件开发工程师网页与数字界面设计师网页开发工程师

3 个职业分类 · 已分类 100%

skill

职业分类

描述

更新

figma-bridge

网页与数字界面设计师

Sync Figma context into implementation plans.

原文语言：英语

2026-03-31

figma-sync

网页开发工程师

Sync a Figma-driven UI task into production code while preserving project conventions and validating visual parity.

原文语言：英语

2026-03-31

git-release

软件开发工程师

Draft release notes, propose a version bump, and keep release mechanics consistent for Claude Code workflows.

原文语言：英语

2026-03-31

git-release

软件开发工程师

Create consistent releases and changelogs.

原文语言：英语

2026-03-31

已展示 4 / 4 个已收集 Skill。