Axon CLI: Git-Based AI Skill Sync Tool

Axon CLI is an innovative tool using Hub-and-Spoke architecture to help developers sync AI editor skills, workflows, and commands across multiple machines. It leverages Git version control for standardized, collaborative skill management, solving the fragmentation issue of AI tool-specific skill storage.

Key benefits:

• Single source of truth for skills via central Hub
• Cross-device sync using Git push/pull
• Support for 20+ mainstream AI tools
• Full lifecycle management (init, sync, audit, rollback)
• Security scanning and semantic search features

Background

With the rise of AI programming assistants (Cursor, Windsurf, Claude Code, Gemini, etc.), developers rely on 'skills' (Markdown docs defining domain knowledge, workflows, tool specs) to extend AI capabilities. However, each tool stores skills in its own location, leading to:

• No sync across devices →割裂 user experience
• Repeated configuration → wasted time

Axon CLI addresses this pain point by unifying skill storage and syncing.

Hub-and-Spoke Architecture

Axon uses a classic Hub-and-Spoke model:

• Hub: Central Git-managed repo at ~/.axon/repo/ with standard directories:
  • skills/: Reusable AI skills
  • workflows/: Scenario-specific workflows
  • commands/: Executable scripts
• Spokes: Symbolic links from AI tool skill directories to the Hub. Example links:
  • `~/.codeium/windsurf/skills → ~/.axon/repo/skills/
  • `~/.cursor/skills → ~/.axon/repo/skills/

All tools share the same skill library—updates in Hub reflect immediately in linked tools, and Git sync keeps multi-device environments consistent.

Key Features

Axon provides comprehensive commands for skill management:

• Init & Link: axon init creates Hub (clone remote or start fresh); axon link creates symlinks, backing up existing content to ~/.axon/backups/.
• Sync: axon sync runs Git add/commit/pull/push. Supports read-write (default) and read-only modes.
• Rollback: axon rollback reverts to history versions safely (no Git knowledge needed, creates new commits).
• Audit: axon audit uses AI (GPT-4o-mini default) to scan for security risks (hardcoded keys, shell injection, PII) with risk grading and auto-fix.
• Search: axon search supports keyword and semantic search (with OpenAI embeddings for vector similarity).
• Vendor Sync: axon vendor sync imports from external GitHub repos without submodules, keeping Hub history clean.

Usage Modes

Axon supports three modes:

1. Pure Local: axon init (no remote repo) → single machine use.
2. Personal Remote: axon init <git-url> or set remote later → sync across multiple machines (most common).
3. Public Upstream: axon init --upstream (clone official kamusis/axon-hub) → read-only access to community-curated skills.

Choose the mode that fits your workflow!

Technical Details

• Cross-Platform: Works on Windows (needs admin for symlinks), WSL (no restrictions). axon init auto-sets .gitattributes to use LF line endings.
• Conflict Resolution:
  • Same MD5 files → duplicate, keep one.
  • Same name different content → add source suffix (e.g., .conflict-antigravity.md).
• Submodule Handling: Strips nested .git folders from downloaded skills before commit.

These details ensure smooth operation across environments.

Supported AI Tools

Axon integrates with 20+ tools. Key support table:

Tool	Skill Support	Workflow Support	Command Support
Windsurf	✓	✓
Cursor	✓
Claude Code	✓		✓
Gemini	✓		✓
VSCode	✓
OpenClaw	✓
Codex	✓
Trae	✓

This wide support unifies skill management across your favorite AI tools.

Practice Tips & Conclusion

Practice Suggestions:

1. Use descriptive skill names (e.g., python-debugging, react-hooks-guide).
2. Write clear commit messages for skill updates.
3. Run axon audit before pushing to public repos to avoid sensitive info leaks.
4. Enable semantic search when skill count exceeds 20.
5. Contribute to community repos via PRs.

Conclusion: Axon CLI paves the way for standardized, collaborative AI skill management. It’s more than a sync tool—it’s a foundation for future AI skill ecosystems, making development environments consistent and efficient for all.