The skill composer for AI agents — compose, diff, and merge skills from any source with AI-powered conflict resolution
Project description
skillhub
There are 50 tools to install AI agent skills. There is exactly one tool that merges them.
skillhub is the skill composer — merge skills from any source across 13 ecosystems (Anthropic, OpenAI, Copilot, Microsoft, Google, Vercel, Agency Agents, and more) into one expert skill, with AI-powered conflict resolution.
pip install skillhub-ai
The problem
You install 6 skills across Claude Code, Cursor, and Codex. Now you have:
- The same
## When to Usesection in 4 different files, all contradicting each other - A
python-patternsskill that conflicts with yourCLAUDE.md security-reviewin Claude but not in Cursor- No way to combine two skills without manually copy-pasting markdown
skillhub solves all of this.
See what's in two skills before you merge them
skillhub diff python-patterns react-patterns
Section-by-section comparison. Green = unique to A. Yellow = conflict. Blue = unique to B. You know exactly what will happen before you compose.
Compose — the feature nobody else has built
skillhub compose python-patterns security-review api-design -o fastapi-expert
Three skills become one expert. All sections merged. Conflicts detected and resolved. One file written to your agent.
# In Claude Code:
/fastapi-expert
Your agent now applies Python conventions, OWASP security, and REST best practices in every response — without you prompting for each one.
AI-powered merge — Claude resolves conflicts intelligently
pip install "skillhub-ai[ai]" # adds the anthropic SDK
export ANTHROPIC_API_KEY=sk-...
skillhub compose python-patterns security-review -o secure-python --strategy ai
Two strategies:
| Strategy | What happens on conflict |
|---|---|
first-wins (default) |
First skill's section wins, conflict logged |
ai |
Claude reads both versions and writes a unified best-of-both |
When --strategy ai is set, conflicting sections like ## Error Handling from two different skills get merged by Claude into a single, coherent section that preserves the best guidance from both.
Pre-built expert templates
skillhub templates # list all 5
skillhub compose --template fastapi-expert # instant expert
skillhub compose --template fastapi-expert --strategy ai # with AI merge
| Template | Skills inside | What you get |
|---|---|---|
fastapi-expert |
python-patterns + security-review + api-design | FastAPI backend expert |
fullstack-expert |
react-patterns + python-patterns + api-design | Full-stack expert |
ml-platform |
agent-builder + mle-workflow + llm-evaluator | ML platform engineer |
pre-pr-reviewer |
code-reviewer + security-review + test-writer | Pre-PR quality gate |
research-analyst |
research-agent + rag-evaluator + doc-generator | Research analyst |
Compose from any source — every skill ecosystem on GitHub
# Merge an official Anthropic skill with an OpenAI skill and your local standards
skillhub compose anthropic:claude-api openai:aspnet-core ./my-team-standards.md \
-o cloud-ai-expert
# Mix GitHub Copilot + Microsoft official + Google Cloud skills
skillhub compose copilot:acquire-codebase-knowledge \
microsoft:skill-creator \
google:agent-platform-deploy \
-o enterprise-expert
# Community skills from every corner of GitHub
skillhub compose agency-agents:frontend-developer \
addyosmani:code-review-and-quality \
scientific:astropy \
antigravity:ab-test-setup \
-o community-expert
# Gamedev skills across all engines
skillhub compose gamedev:godot-best-practices \
skills.sh:react-expert \
python-patterns \
-o gamedev-web-expert
# Use the raw github: escape hatch for anything not yet indexed
skillhub compose github:owner/repo/path/to/SKILL.md python-patterns -o custom-expert
| Source prefix | Ecosystem | GitHub repo |
|---|---|---|
name (no prefix) |
skillhub registry | chandrudp29/skillhub |
anthropic:name |
Anthropic official skills | anthropics/skills |
openai:name |
OpenAI / Codex official skills | openai/skills |
copilot:name |
GitHub Copilot awesome skills | github/awesome-copilot |
microsoft:name |
Microsoft official skills | microsoft/skills |
google:name |
Google official skills | google/skills |
skills.sh:name |
Vercel skills.sh registry | vercel-labs/skills |
agency-agents:name |
Agency Agents role personas | msitarzewski/agency-agents |
addyosmani:name |
Addy Osmani production skills | addyosmani/agent-skills |
scientific:name |
K-Dense AI scientific skills | K-Dense-AI/scientific-agent-skills |
antigravity:name |
Antigravity / OpenClaw community | sickn33/antigravity-awesome-skills |
gamedev:name |
Game Dev skills | gamedev-skills/awesome-gamedev-agent-skills |
tech-leads:name |
Tech Leads Club validated skills | tech-leads-club/agent-skills |
claude:name |
Already in your Claude Code project | .claude/commands/ |
cursor:name |
Already in your Cursor project | .cursor/rules/ |
codex:name |
Already in your Codex project | AGENTS.md blocks |
gemini:name |
Already in your Gemini CLI project | .gemini/skills/ |
github:owner/repo/path |
Any public GitHub file | raw.githubusercontent.com |
./path/to/file.md |
Local file on disk | — |
Install skills — works for all 4 agents
skillhub install debug-agent # Claude Code (default)
skillhub install debug-agent --all-agents # all 4 agents at once
skillhub install debug-agent --dry-run # preview before writing
| Agent | Install Path |
|---|---|
| Claude Code | .claude/commands/<name>.md |
| Cursor | .cursor/rules/<name>.mdc |
| OpenAI Codex | AGENTS.md (appended with markers) |
| Gemini CLI | .gemini/skills/<name>.md |
Search and discover
skillhub list # browse all 26 skills
skillhub search "security" # search by keyword
skillhub search "react" --tag ui # filter by tag
skillhub info debug-agent # full preview before installing
Create your own skill
skillhub init our-coding-standards
# → creates our-coding-standards/SKILL.md with template
# Edit it, then:
skillhub install our-coding-standards # test locally → /our-coding-standards in Claude
skillhub publish our-coding-standards/ # open a PR to add it to the registry
How compose works
skillhub compose python-patterns security-review api-design -o fastapi-expert
│
▼
┌──────────────────────────────────────────┐
│ Fetch each skill from its source │
│ registry / skills.sh / GitHub / local │
└──────────────────────────────────────────┘
│
▼
┌──────────────────────────────────────────┐
│ Parse all ## sections │
│ Detect conflicts across skills │
└──────────────────────────────────────────┘
│
├─── strategy: first-wins ──→ first skill's section kept
│
└─── strategy: ai ─────────→ Claude merges both intelligently
│
▼
┌──────────────────────────────────────────┐
│ Write composed skill to agent path │
│ .claude/commands/fastapi-expert.md │
│ │
│ Updates .claude/skills.json index │
└──────────────────────────────────────────┘
│
▼
/fastapi-expert ✓
Full command reference
| Command | What it does |
|---|---|
skillhub diff <a> <b> |
Compare two skills section by section |
skillhub compose <a> <b> -o <name> |
Merge skills (first-wins on conflicts) |
skillhub compose <a> <b> -o <name> --strategy ai |
Merge with Claude AI resolution |
skillhub compose --template <name> |
Use a pre-built expert template |
skillhub templates |
List all built-in templates |
skillhub search <query> |
Search by name, description, or tag |
skillhub list |
Browse all 26 skills |
skillhub list --installed |
Show skills installed in this project |
skillhub info <name> |
Full details and preview |
skillhub install <name> |
Install for Claude Code |
skillhub install <name> --all-agents |
Install for all 4 agents |
skillhub install <name> --dry-run |
Preview without writing files |
skillhub install <name> --overwrite |
Replace existing installation |
skillhub uninstall <name> |
Remove a skill |
skillhub update |
Update all installed skills to latest |
skillhub init <name> |
Scaffold a new SKILL.md template |
skillhub publish <path> |
Submit to the registry via GitHub PR |
Available Skills (26)
Research & Analysis
| Skill | Description |
|---|---|
research-agent |
Multi-source deep research with synthesis and citations |
rag-evaluator |
Evaluate RAG pipelines: faithfulness, relevance, hallucination |
llm-evaluator |
Build evaluation frameworks for LLM outputs |
Code Quality
| Skill | Description |
|---|---|
code-reviewer |
Five-axis code review (correctness, security, performance, readability, maintainability) |
debug-agent |
Systematic root cause analysis — feedback loop first |
refactor-agent |
Safe, incremental refactoring with verification |
test-writer |
Write meaningful tests that actually catch bugs |
security-review |
OWASP-based security auditing |
Documentation
| Skill | Description |
|---|---|
doc-generator |
README, API docs, docstrings, changelogs |
pr-summarizer |
Generate PR descriptions: what, why, how, risks |
Development Patterns
| Skill | Description |
|---|---|
python-patterns |
Modern Python best practices (typing, async, dataclasses) |
react-patterns |
React 18+ patterns (hooks, suspense, server components) |
fastapi-patterns |
Production FastAPI patterns |
api-design |
REST API conventions and best practices |
sql-agent |
Write, optimize, and explain SQL with query plans |
DevOps
| Skill | Description |
|---|---|
docker-agent |
Dockerfile optimization, multi-stage builds |
git-workflow |
Branching strategies, commit conventions |
AI / ML Engineering
| Skill | Description |
|---|---|
agent-builder |
Build LangGraph agents: tools, memory, streaming, multi-agent |
prompt-optimizer |
Diagnose and improve underperforming prompts |
mle-workflow |
Production ML workflow: data, training, deployment |
cost-tracker |
Track and optimize LLM API costs |
Career
| Skill | Description |
|---|---|
yc-job-tracker |
Daily AI/ML job tracking at YC startups |
career-ops |
Global AI/ML job hunt — USA, Europe, Singapore, Dubai |
Community (via agency-agents)
| Skill | Description |
|---|---|
frontend-developer |
Senior frontend developer expertise |
backend-architect |
Senior backend architecture and system design |
penetration-tester |
Security testing and penetration testing methodology |
Quick start
pip install skillhub-ai
# Browse what's available
skillhub list
skillhub templates
# See what's in two skills before merging
skillhub diff debug-agent code-reviewer
# Compose a pre-built expert (5 templates available)
skillhub compose --template fastapi-expert
# Compose your own
skillhub compose debug-agent code-reviewer -o my-reviewer
# Use AI to resolve conflicts intelligently
pip install "skillhub-ai[ai]"
export ANTHROPIC_API_KEY=sk-...
skillhub compose debug-agent code-reviewer -o my-reviewer --strategy ai
# Pull from any source
skillhub compose skills.sh:react-expert python-patterns -o my-expert
Cookbook
Copy-paste recipes for the most common workflows → docs/cookbook.md
Contributing
- Fork this repo
- Add your skill under
skills/<name>/SKILL.md - Add an entry to
registry/index.json - Open a PR — we review for usefulness, not quantity
Or use the CLI:
skillhub init <name>
# edit the SKILL.md
skillhub publish <name>/
Author
Built by Chandrashekar DP — Senior AI/ML Engineer.
If skillhub saves you time, a ⭐ helps others find it.
License
MIT
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file skillhub_ai-0.3.0.tar.gz.
File metadata
- Download URL: skillhub_ai-0.3.0.tar.gz
- Upload date:
- Size: 1.5 MB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.11
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
fe52621fe486f371a235e68ab2f50d7be4dc8df7d11932d1244d05221db98830
|
|
| MD5 |
a3ac791f2d700adcf3b682974df65281
|
|
| BLAKE2b-256 |
42867187a4e33e7c2ebe5df7044b92ceee724da9b5e356bcc5673af7af9b2954
|
File details
Details for the file skillhub_ai-0.3.0-py3-none-any.whl.
File metadata
- Download URL: skillhub_ai-0.3.0-py3-none-any.whl
- Upload date:
- Size: 25.3 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.2.0 CPython/3.12.11
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
cd05902d36eb293db6bee7625eb26abe86d896d154112e48ff1408c6c3778269
|
|
| MD5 |
c0d62e96b74bc3b62312673412b14ef0
|
|
| BLAKE2b-256 |
58ae9fa06216b6cdb16152804d37903603e095a4dd622f835f47fdadad6b3f9e
|