Reusable knowledge plugins for AI agents. Think of it as npm for AI agent skills — structured knowledge modules that any AI system can consume to learn new capabilities.
🎯 Not locked to any framework. These plugins work with GitHub Copilot, Claude, ChatGPT, Squad, or any LLM-based agent. A plugin is just a well-structured markdown file that teaches an AI agent how to do something.
AI agents are powerful, but they don't know everything out of the box. This marketplace provides drop-in knowledge plugins — each one teaches your AI agent a specific skill:
- 🖥️ Automate Microsoft Teams UI when the Graph API isn't enough
- 📡 Bridge Teams messages into GitHub issues automatically
- 🔄 Coordinate work across machines using Git as a task queue
- 📋 Manage project boards with proper lifecycle transitions
- 🤝 Distribute tasks across agents without conflicts
Each plugin is a self-contained SKILL.md — structured markdown that any AI agent can read and act on. No SDKs, no dependencies, no lock-in.
| Plugin | Description | Triggers |
|---|---|---|
| 🖥️ teams-ui-automation | Hybrid Teams automation — Playwright + keyboard shortcuts + UIA | teams ui, install teams app, teams automation |
| 🔐 bitwarden-credential-management | Bitwarden MCP + CLI integration for secure AI agent credential management | bitwarden, vault, credential, password manager |
| 📡 teams-monitor | Monitor Teams channels via WorkIQ and bridge messages to GitHub issues | check teams, teams monitor |
| 🔄 cross-machine-coordination | Git-based task queuing for multi-machine agent coordination | cross machine, remote task, distribute work |
| 📋 github-project-board | GitHub Projects V2 board sync with issue lifecycle | project board, move issue, board status |
| 🤝 github-distributed-coordination | Distributed work claiming protocol using GitHub-native features | distributed coordination, work claiming |
| 📧 outlook-automation | Control Outlook on Windows — send emails, create meetings, search inbox, manage tasks | create meeting, send email, search emails, outlook |
| 📰 news-broadcasting | Tech news scanning, compilation, and delivery to team channels | news, tech news, daily briefing, news report, scan news |
| ✅ fact-checking | Agent fact-verification patterns for ensuring accuracy before publishing | fact check, verify, validate claims, source check |
| 🔄 session-recovery | Find and resume individual past Copilot CLI sessions from session_store | session recovery, find session, resume session |
| 🔄 restart-recovery | Snapshot and restore full dev environment (services, sessions, state) after machine restart | restart recovery, recover from restart, snapshot before restart |
| 🔄 reflect | Agent self-reflection and continuous improvement patterns | reflect, retrospective, self-improve, lessons learned, what went wrong |
| 🔀 github-multi-account | Solve multi-account GitHub CLI chaos with account-locked aliases | gh auth switch, wrong account, multi account, EMU |
| 📱 phone-link-2fa | Extract 2FA codes from SMS via Windows Phone Link | 2fa, sms, verification code, otp, phone link |
| 📘 conference-book-of-news | Auto-generate Ignite-style "Book of News" PDF from any conference — scrapes videos, captures screenshots, OCR extracts slides, AI summarizes sessions | conference, book of news, session recap, ignite, build, summit, video screenshots |
| 🚀 spawn-squad | Teach an HQ agent to create, run, and manage child squads — fan-out pattern for parallel exploration, legacy migration, and ephemeral missions | spawn squad, create squad, fan out, child squad, new squad, delegate squad, launch squad, spin up squad, orchestrate squads |
| 🤝 cross-squad-communication | Protocol for querying, delegating, and sharing context across independent Squad instances in other repos | cross squad, inter squad, squad handoff, delegate to squad |
| 🧠 governed-memory-cli-bridge | Workaround for Squad 0.10.x — teaches agents to invoke squad memory classify/write/search/audit via their shell tool, because the governed memory tools aren't yet bridged through the squad_state MCP server |
governed memory, memory cli, memory classify, memory write, memory search, memory audit, audit trail, forbidden scan, loadGuidance, workaround |
Every plugin's core is a single file: SKILL.md. The fastest way to install plugins is with the Copilot CLI plugin system.
# One-time: register the marketplace
copilot plugin marketplace add tamirdresher/squad-skills
# Install any plugin
copilot plugin install teams-ui-automation@squad-skills
# Or install directly from the repo (no marketplace registration needed)
copilot plugin install tamirdresher/squad-skills:plugins/teams-ui-automationPlugins are installed to ~/.copilot/state/installed-plugins/ and activate automatically. Manage them with:
copilot plugin list # View installed plugins
copilot plugin update PLUGIN-NAME # Update to latest version
copilot plugin uninstall PLUGIN-NAME # Remove plugin💡 In VS Code Copilot Chat,
/plugin installworks in interactive sessions too.
# Copy the entire plugin folder — Squad agents auto-discover from .squad/skills/
cp -r plugins/teams-ui-automation /path/to/your/repo/.squad/skills/- Open your Claude Project settings → Project Knowledge
- Paste the contents of
SKILL.mdas a knowledge document
- Go to Configure → Knowledge
- Upload the
SKILL.mdfile
Just read the SKILL.md file. That's it. It's structured markdown with:
- YAML frontmatter — metadata, triggers, confidence level
- Recipes — step-by-step procedures the agent can follow
- Error recovery — what to do when things go wrong
- Anti-patterns — common mistakes to avoid
# Example: Load a plugin in your custom agent
with open("plugins/teams-ui-automation/SKILL.md") as f:
skill_knowledge = f.read()
# Add to your agent's system prompt or context
agent.add_context(skill_knowledge)New to Squad Skills? These guides will get you up and running:
- Getting Started Guide — New here? Walk through picking, installing, and using your first plugin in under 5 minutes.
- Plugin Selection Guide — Not sure which plugin you need? Browse the categorized matrix with complexity ratings and platform info.
Each plugin follows a consistent structure:
plugins/
your-plugin/
manifest.json # 📋 Machine-readable metadata
SKILL.md # 🤖 Agent-consumable knowledge (the core)
README.md # 📖 Human-readable documentation
scripts/ # 🔧 Supporting scripts (optional)
Every plugin includes a manifest.json for tooling and discovery:
{
"name": "your-plugin-name",
"version": "1.0.0",
"description": "What it does in one line",
"platforms": ["copilot", "claude", "squad", "any"],
"triggers": ["keyword1", "related phrase"],
"author": "Your Name",
"license": "MIT",
"files": {
"skill": "SKILL.md",
"scripts": ["scripts/helper.ps1"]
}
}We'd love your plugins! If you've built an automation pattern that works well with AI agents, share it.
Quick start:
- Fork this repo
- Create
plugins/your-plugin-name/withmanifest.json,SKILL.md, andREADME.md - Open a PR
📖 See the full Contributing Guide for templates, best practices, and quality guidelines.
- Knowledge, not code. Plugins are structured documentation, not libraries. They teach agents how to do things, not what to execute.
- Platform-agnostic. If it can read markdown, it can use these plugins.
- Self-contained. Each plugin has everything an agent needs — context, recipes, error recovery, and anti-patterns.
- Community-driven. The best automation patterns come from real-world use.
MIT — use these plugins however you want.