Skip to content

asleekgeek/claude-multiagent-catalogue

Repository files navigation

Claude Multiagent Catalogue

A Claude Code catalogue mirroring the awesome-copilot marketplace — agents, instructions, skills, plugins, hooks, workflows, and cookbook recipes, ported into Claude Code-native form.

The repo name is legacy. Scope is not limited to multi-agent rigs — it covers every artefact type upstream publishes. The "multiagent" portion of the name reflects the original (narrower) intent; see CLAUDE.md for the full scope correction.

Every artefact is a markdown agent definition, SKILL.md, instructions file, hook bundle, workflow prompt, or cookbook recipe — no application code.

What's in this repo

Bucket Path Status Upstream
🤖 Agents agents/ 211 / 211 awesome-copilot/agents/
📋 Instructions instructions/ 183 / 183 awesome-copilot/instructions/
🎯 Skills skills/ 335 / 338 awesome-copilot/skills/
🔌 Plugins plugins/ 67 / 67 awesome-copilot/plugins/
🪝 Hooks hooks/ 6 / 6 awesome-copilot/hooks/
Workflows workflows/ 8 / 8 awesome-copilot/workflows/
🍳 Cookbook cookbook/ 7 / 10 (3 skipped-with-reason) awesome-copilot/cookbook/
🛠️ Conversion skill copilot-to-claude-multiagent/ (this repo)

The conversion procedure for every bucket lives in copilot-to-claude-multiagent/SKILL.md.

Quickstart

# See what's available across every bucket
./install.sh --list

# Install a plugin's agents into the current project
cd ~/my-project
/path/to/claude-multiagent-catalogue/install.sh rug-agentic-workflow

# Install a plugin + its skills
/path/to/claude-multiagent-catalogue/install.sh ai-team-orchestration --skills

# Install a single standalone agent (Phase B+)
/path/to/claude-multiagent-catalogue/install.sh install-agent csharp-expert

# Install a single instruction (Phase D+)
/path/to/claude-multiagent-catalogue/install.sh install-instruction typescript

# Install a hook bundle (Phase E+)
/path/to/claude-multiagent-catalogue/install.sh install-hook secrets-scanner

# Verify
ls .claude/agents/
ls .claude/skills/
ls .claude/instructions/

Agents land in .claude/agents/, skills in .claude/skills/, instructions in .claude/instructions/, hook scripts in .claude/hooks/<bundle>/. Claude Code discovers agents and skills automatically; instructions and hooks require explicit wiring (see each bucket's README for details).

Plugins (currently populated)

All 67 upstream plugins are mirrored — 5 multi-agent orchestration rigs (full detailed entries in CATALOGUE.md) plus 62 topic bundles (languages, clouds, SaaS platforms, tools).

The 5 multi-agent rigs:

Plugin Pattern Components
rug-agentic-workflow Orchestrator + SWE + QA subagents via the Task tool (RUG — Repeat Until Good) agents
ai-team-orchestration Producer / Dev / QA across parallel Claude Code sessions, human as message bus agents + skill
software-engineering-team Pool of 7 SDLC reviewer specialists (architect, security, devops, PM, tech writer, UX, responsible-AI) agents
structured-autonomy Plan → Generate → Implement skill pipeline with checkpoints between commits skills
quality-playbook 7-phase quality engineering audit; orchestrator spawns one Task sub-agent per phase agents + skill

The 62 topic bundles cover languages (.NET, TypeScript, Python, JVM, Rust, Go, Swift, etc.), clouds (Azure, AWS), SaaS platforms (Salesforce, Power Platform, Sitecore, Drupal), database admin, CI/CD, testing, security, and AI/ML tooling. See CATALOGUE.md → Topic bundles for the full table or ./install.sh --list for a live listing.

Full plugin entries — orchestration flow, agent rosters with model assignments, install notes — live in CATALOGUE.md.

Repository layout

agents/                       Standalone Claude Code agent files
instructions/                 Coding-standard fragments with applyTo metadata
plugins/<plugin-name>/        Installable plugin (agents/ and/or skills/ + plugin.json + README.md)
skills/                       Standalone skill pool — see skills/STATUS.md
hooks/<bundle-name>/          Hook bundles + claude-settings.json fragments
workflows/                    Agentic GH Actions — <name>.md prompts + .github-action.yml runners
cookbook/                     Translation guides for upstream Copilot-SDK recipes
copilot-to-claude-multiagent/ Conversion skill (SKILL.md + references/)
install.sh                    Installer (handles every bucket above)
CATALOGUE.md                  Discovery index — every populated bucket gets a section
CLAUDE.md                     Repo-level instructions for Claude Code sessions

Conversion procedure

Every upstream artefact is ported through copilot-to-claude-multiagent/SKILL.md. The skill handles all seven bucket types via a Phase 0 dispatcher that branches on the upstream path. In short:

  1. Identify the artefact type from its upstream path.
  2. Apply the matching pattern from copilot-to-claude-multiagent/references/conversion-patterns.md.
  3. Output to the corresponding top-level directory here.
  4. Catalogue — update CATALOGUE.md and the bucket README's status table.

Common transformations applied across buckets: drop the Copilot tools: key, drop the agents: key, map the Copilot/OpenAI model to a Claude tier, rewrite runSubagentTask, #context7 → Context7 MCP, VS Code UI → bash equivalents.

Skill pool

skills/ is a copy-paste surface, not part of the installer. See skills/STATUS.md for the adaptation audit. To use one:

mkdir -p .claude/skills
cp -r /path/to/claude-multiagent-catalogue/skills/<skill-name> .claude/skills/

License and attribution

Each artefact retains its upstream licence — see the per-plugin README.md and any LICENSE.txt shipped alongside the skill. Where a port involved non-trivial adaptation, the file's ## Claude Code Notes section documents what was changed and why.

Issues and contributions

Bug reports and contribution proposals: see the templates in .github/ISSUE_TEMPLATE/.

About

A Claude Code adjusted version of awesome-copilot

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages