agent-skills-bridge
Skill0 starsChanged 4 months agoMITView on GitHub
Cross-Agent Skills Bridge. Discovers and indexes skill directories from 36 AI coding agents (Claude Code, Cursor, GitHub Copilot, Cline, Roo Code, Windsurf, and more) per the Agent Skills open standard (skills.sh). Builds a unified searchable index, detects cross-agent duplicates, and recommends the best version. Loads skills on-demand from original paths — never copies. Trigger: skill bridge, cross-agent skills, find skills, index skills, search skills, 技能桥接, 跨工具技能, 找技能, 搜索技能.
1 other project uses a version of this.
Adopted by
Most recently taken up by
- sethliao/one-person-studiotheir own version, first seen Oct 20, 2026
Other versions1 project uses a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 1
- sethliao/one-person-studio★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 91
roots: Check configured roots paths first
- Distinctive: line 146
Don't re-read: after reading recommended version, skip duplicates
- Distinctive: line 191
Read-only: never modify other agents' skill files
- Very specific: line 196
No hardcoded paths: all paths managed via config.json, code uses only env vars
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Agent Skills Bridge (v3.0)
Universal cross-agent skills index. Discovers skill directories from 36 AI coding agents on your machine, builds a unified searchable index, detects duplicates, and recommends the best version — all from original paths, no copying.
Core Principle: Index, Don't Copy
Don't move other agents' skills into your own library. Build a unified index, load on demand from the original path, done.
| Principle | Detail |
| Index | 37 agents × 2000+ skills indexed, one search covers all |
| Load on demand | When a skill matches, Read its original SKILL.md, follow instructions, move on |
| Dedup | Same skill across multiple agents → auto-recommend best version |
| No blind copying | Only integrate into your own library after ≥3 uses + confirmed superiority |
| Read-only | Never modify other agents' skill files |
What's New in v3.0
| Change | Detail |
| Universal | Expanded from 12 to 36 agents per the Agent Skills open standard |
| Cross-platform | Forward-slash paths, ${HOME} / ${APPDATA} env vars — works on Windows, macOS, Linux |
| No hardcoded paths | Zero machine-specific paths in code or default config |
| Dynamic search_roots | Iterates all config keys (home/config/local) instead of hardcoded keys |
| Bundled config | Full 37-agent config.json ships with the repo |
Supported Agents (37)
Based on the Agent Skills open standard:
| Priority | Agents |
| P1 (major) | Claude Code, Cursor, GitHub Copilot |
| P2 (established) | Cline, Roo Code, Windsurf, Continue, Codex, WorkBuddy |
| P3 (growing) | Gemini CLI, CodeBuddy, Qwen Code, Trae, Trae CN, OpenClaude, Amp, Kimi Code CLI, Goose, Kilo Code, OpenHands, OpenCode |
| P4 (niche) | Antigravity, Moltbot, Command Code, Crush, Droid, Junie, Kiro CLI, Kode, MCPJam, Mux, Pi, Qoder, Zencoder, Neovate, Pochi |
Each agent has standard roots (global path) and discover patterns (project-level path) defined in config.json.## Configuration
config.json
Path: agent-skills-bridge/config.json
{
"version": "3.0.0",
"standard": {
"name": "Agent Skills Open Standard",
"url": "https://github.com/scotttrinh/skills"
},
"search_roots": {
"home": ["${HOME}", "${USERPROFILE}"],
"config": ["${HOME}/.config", "${APPDATA}"],
"local": ["${HOME}/.local", "${LOCALAPPDATA}"]
},
"tools": {
"claude-code": {
"label": "Claude Code",
"roots": ["${HOME}/.claude/skills"],
"fallback": ["${HOME}/.claude/plugins"],
"priority": 1,
"discover": [".claude/skills"]
},
// ... 36 more agents
},
"domain_keywords": { ... }
}Environment Variables
| Variable | Expands to | Platform |
${HOME} | User home directory | All |
${USERPROFILE} | Windows user directory | Windows |
${APPDATA} | AppData Roaming | Windows |
${LOCALAPPDATA} | AppData Local | Windows |
~ | User home directory | All (shorthand) |
Path Validation (3-tier)
- roots: Check configured
rootspaths first - fallback: Try
fallbackpaths if roots don't exist - discover: Auto-search under
search_rootsusingdiscoverpatterns
Discovered paths are saved back to config.json automatically.
Workflow
DISCOVER → VALIDATE → SCAN → INDEX → DEDUP → CACHE → SEARCH → MATCH → LOAD → EXECUTEQuick Start
# Validate which agents are installed on your machine
python scripts/build_index.py --validate
# Auto-discover agent paths and update config
python scripts/build_index.py --discover
# Build the full index (scan all agent skill dirs)
python scripts/build_index.py --print
# Search for skills by keyword
python scripts/build_index.py --search "code review"
# Search within a specific agent
python scripts/build_index.py --search "patent" --tool claude-code
# List all skills from a specific agent
python scripts/build_index.py --list --tool cursor
# Show duplicate analysis (which skills exist across multiple agents)
python scripts/build_index.py --duplicatesDedup Analysis
Quality Scoring (7 dimensions)
| Dimension | Score | Detail |
| description non-empty | +10 | Has documentation |
| description > 50 chars | +5 | Detailed |
| description > 100 chars | +5 | Very detailed |
| keywords > 5 | +10 | Well-indexed |
| dir name = frontmatter name | +10 | Properly named |
| TODO/WIP in description | -5 | Unfinished |
| Tool priority bonus | +15/+10/+5/+0 | P1/P2/P3/P4 |
When to call recommended version
- Prefer
[推荐]version: highest quality score - Call non-recommended only when: recommended version lacks a feature you need, or its agent isn't available in current environment
- Don't re-read: after reading recommended version, skip duplicates
Integration Criteria
Not every good skill should be copied. Only integrate when ALL conditions are met:
| Condition | Detail |
| ✅ Repeated use | Called ≥ 3 times in recent work |
| ✅ Better than own | Clearly superior to existing skills |
| ✅ Self-contained | No dependency on source agent's proprietary commands |
| ✅ Improvable | Can be improved based on usage experience |
Task Matching
| Trigger | Search scope | Keywords |
| Research/academic | Claude Code > Hermes > CC Switch | paper/research/论文 |
| Novel writing | Hermes > CC Switch | novel/editing/小说 |
| Engineering | Claude Code > OpenClaw > .agents | code/debug/test |
| Patent | CC Switch > Hermes | patent/专利/交底 |
| Data visualization | CC Switch > Claude | matplotlib/figure/图表 |
| Document processing | Claude Code > OpenClaw | docx/xlsx/pdf |
| Fund application | Hermes > Claude Code | nsfc/fund/基金 |
| Unknown domain | All agents | keyword search first |
Migration Guide
To a new machine
- Copy the
agent-skills-bridge/directory to the new machine - Run
python scripts/build_index.py --discover— auto-detects installed agents - Check
python scripts/build_index.py --validate— see which agents were found - Build
python scripts/build_index.py --print— generate the index - Optional: manually edit
config.jsonif an agent is installed in a non-standard location
Share with others
- Package the
agent-skills-bridge/directory (excludeskills_index.json— it's a machine-specific cache) - The bundled
config.jsonuses${HOME}env vars — auto-adapts to any machine - Recipient runs
--discoverto find their installed agents
Notes
- Index, not copy: the purpose is unified search + on-demand loading, not skill migration
- Read-only: never modify other agents' skill files
- Skills are documents: SKILL.md files are AI behavioral instructions, not executables
- Script safety: evaluate scripts/ in skills before running (read source, check for destructive ops)
- Encoding: some older SKILL.md files may use GBK encoding; the parser falls back to GBK/latin-1
- Graceful degradation: missing agent directories are skipped, don't affect others
- No hardcoded paths: all paths managed via config.json, code uses only env vars
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

