Skills
Repeatable workflows for agents — how your team works.
Shared Agents
Team skills and learnings for AI assistants
One install, Git-synced, IDE-agnostic. Give Cursor, Claude Code, Zed, Codex, and the rest the same skills, rules, and team memory — without locking into a single vendor.
Speed
Top Performance
Privacy First
Team data stays private
Developer Experience
Exceptional
Tool Agnostic
Works everywhere
Why this exists
Teams use many AI tools. Each session starts cold — no shared workflows, no institutional memory. Shared Agents is one place for skills, rules, and learnings; everyone syncs the same content, and sensitive team data stays in your private repo.
Rules
Shared instructions per tool (Cursor .mdc, AGENTS.md, CLAUDE.md).
Learnings
Institutional memory: bugs, stack quirks, decisions — reviewed by humans.
Get started
Install in seconds
Copy the command, run it in your terminal, then type sa to sync skills, rules, and learnings across your team.
install — shared-agents
- 1
Run installer
Core, team repo, and tool adapters in one wizard.
curl … | bash - 2
Reload shell
Load PATH and the sa command.
source ~/.bashrc - 3
Verify
Check adapters and symlinks.
sa check
How it works
One workflow for the whole team
Install once, then sync, share, and review team knowledge across every tool. Command details: CLI reference.
Sync
Pull the latest Core and team content. Skills and rules link into every IDE automatically.
CLI reference →Share
Team skills and rules live in your private repo — same workflows in Cursor, Claude Code, Zed, and more.
Skills & rules →Review
Agents propose learnings; humans approve with sa review. Only approved knowledge reaches the team.
Learnings →Explore
Installation, adapters, paths, and troubleshooting — everything in the docs.
All guides →