team.management vs claude-code-setup.
claude-code-setup is Anthropic’s plugin in the official marketplace: it reads your codebase and recommends the hooks, skills, MCP servers, and subagents it could use. team.management is the other half of that sentence — the engine that makes the process behind them hold at runtime.
What claude-code-setup is
claude-code-setup is Anthropic’s own entry in the official plugin marketplace — in its words, it exists to “analyze codebases and recommend tailored Claude Code automations such as hooks, skills, MCP servers, and subagents.” Under the hood it is one skill, claude-automation-recommender, plus five reference catalogs of common patterns. Ask it what your project should automate and it scans the repo — language, framework, tests, CI — and returns a report with the top one or two recommendations per category, each with pointers to install or build it.
It is deliberately hands-off: “It does NOT create or modify any files. Users implement the recommendations themselves or ask Claude separately to help build them,” per the skill’s own instructions. As of August 2026 it ships no hooks, no slash commands, no subagents, and no MCP servers of its own — the categories in its report are things it recommends, not things it contains.
What team.management is
team.management is an open-source protocol engine that runs inside Claude Code and picks up exactly where a recommendation stops: it ships the hooks, the subagents, and the workflow as one system, already assembled. Work moves through named protocols; at every step a PreToolUse hook decides, before each tool call, what the agent may do right now (the DAIC loop is the mechanics). Each task gets a branch, a work log, and an audit trail without anyone remembering to ask.
The core difference — a report vs rails
claude-code-setup ends where your work begins: its output is a markdown report, and everything after that is voluntary. Whether the hooks get written, whether the review step actually runs before merge, whether the workflow it sketched is still followed on a Friday afternoon — all of that lives where it always lived: discipline. team.management starts at exactly that line. The engine holds which step you’re in, gates the tools accordingly, and writes down what happened — the difference between being told your project could use guardrails and driving with them bolted down.
And the symmetry: rails bind the agent, not you. The config is MIT, in your own repo — you can strip a gate any time you choose. The model, mid-session, cannot. That asymmetry is the entire point.
| claude-code-setup | team.management | |
|---|---|---|
| Own words | “Analyze codebases and recommend tailored Claude Code automations such as hooks, skills, MCP servers, and subagents.” | A protocol engine that enforces the workflow those automations serve |
| Mechanism | One read-only skill + five reference catalogs; outputs a recommendations report | Protocol engine + PreToolUse hooks gating tools at runtime |
| What changes in your repo | Nothing — “It does NOT create or modify any files” | Task files, branches, work logs, protocol state — engine-managed |
| Workflow lifecycle | None — suggests automations, then exits | Named protocols sequence the steps; skipping isn’t offered |
| Review & QA | May recommend a review subagent | Review is a gated step; Codex and Antigravity can join as reviewers |
| Runs | On demand, when you ask for recommendations | Every session, for the life of the task |
| Repository | github.com/anthropics/claude-plugins-official | github.com/TeamManagementPlugin/claude-plugin |
| Price & license | Free, Apache-2.0 | Free, MIT |
Run the recommender, then bolt down the rails
These two aren’t rivals; they’re consecutive steps. claude-code-setup is a good first move on any repo — a fast, official map of what your project could automate. Read its report, adopt what fits. Then notice which categories keep appearing — hooks, subagents, workflow automations — and install the thing that ships them pre-assembled and enforced. The report tells you what your setup is missing. The rails are how it stays followed.
FAQ
Does claude-code-setup set up hooks and subagents for me?
No — by its own documentation it is a read-only advisor. The skill analyzes your codebase and writes a recommendations report; it does not create or modify any files. You implement the suggestions yourself, or ask Claude to build them in a separate step.
Can claude-code-setup and team.management run together?
Naturally. Run the recommender first — its report maps what your project could automate. Where the report says hooks, subagents, and workflow automations, team.management is that slot filled as one maintained system: named protocols, a task lifecycle, and runtime gates, instead of a list of parts to assemble.
What’s the difference if both involve hooks?
claude-code-setup suggests hooks you could write for your project. team.management ships them as a working engine: a PreToolUse hook runs before every tool call and allows or blocks it based on the protocol step you’re in. One hands you a parts list; the other is the assembled machine.
Facts and figures on this page are as of August 2026, verified against the sources linked inline. If you’re reading this much later — check the sources.