Workspace — the maintainer hub

The agentteamland/workspace repo is the maintainer hub for the AgentTeamLand ecosystem. It's a meta-repo: cloning it and running one script gives you every peer repo (cli, core, registry, software-project-team, etc.) checked out under a single tree at ./repos/. Every moving part of the platform is one cd repos/<name> away.

Use the workspace when you're doing maintenance work that spans multiple repos: cross-repo refactors, multi-PR rollouts, governance audits, or just git status across the whole org without 14 separate cd commands.

If you only want to USE atl (install teams in your own projects), you don't need the workspace — brew install agentteamland/tap/atl is enough. The workspace is for ecosystem-side work.

Bootstrap

git clone https://github.com/agentteamland/workspace.git
cd workspace
./scripts/sync.sh

sync.sh clones every peer repo under agentteamland/ into ./repos/<name>/. It's idempotent — re-running fast-forward-pulls existing clones and clones any new ones added to the org since last run.

After sync, ./repos/ contains the full org snapshot (16 repos as of 2026-05-03):

repos/
├── cli/                       # atl binary (Go) — the CLI users install
├── core/                      # global skills + rules + JSON schemas
├── brainstorm/                # /brainstorm skill + its rule
├── rule/                      # /rule + /rule-wizard skills
├── team-manager/              # bootstrap install.sh (delegates to atl post-v1.0.0)
├── software-project-team/     # 13 agents + 3 skills (.NET + Flutter + React stack)
├── design-system-team/        # 2 agents + 10 /dst-* skills (native design + prototype)
├── starter-extended/          # inheritance example team
├── create-project/            # 🗄 ARCHIVED 2026-05-04 — scaffolder moved into teams; kept for history
├── registry/                  # teams.json — canonical team catalog
├── docs/                      # this docs site (VitePress, EN + TR)
├── homebrew-tap/              # auto-managed by goreleaser
├── scoop-bucket/              # auto-managed by goreleaser
├── winget-pkgs/               # fork of microsoft/winget-pkgs
└── .github/                   # organization profile

Daily commands

The workspace ships three scripts under ./scripts/:

./scripts/sync.sh         # clone missing repos; fast-forward pull existing ones
./scripts/status.sh       # tabular overview — who's dirty, ahead, behind
./scripts/push-all.sh     # dry-run list of unpushed commits (use --force to push)

status.sh prints a one-line-per-repo table — branch, ahead/behind counts, dirty marker. Run it at the start of any session to see the org's current state at a glance.

push-all.sh is dry-run by default — it shows what WOULD push but doesn't actually push. Pass --force to actually push. (The "force" name refers to overriding the dry-run, not git push --force — the actual push uses normal git semantics.)

Working in a peer repo

cd repos/<repo-name>
# Make your changes, follow the team-repo-maintenance discipline
git checkout -b <type>/<short-description>
# ... edit files ...
git add <files> && git commit -m "<conventional message>"
git push -u origin <branch-name>
gh pr create
# Wait for review + merge by the maintainer

Each peer repo is its own git clone with its own remote. Branch protection on every public production repo (12 of the 16 — release-pipeline + .github excluded) enforces the PR flow. See Team-repo maintenance for the full discipline.

Using the workspace with Claude Code

Open Claude Code in the workspace root:

cd ~/projects/my/agentteamland/workspace
claude    # or however you invoke Claude Code

When Claude Code starts here, it automatically sees:

• Every peer repo under ./repos/ for direct editing — no separate cd needed
• All active brainstorms (auto-pinned into CLAUDE.md per the brainstorm rule)
• Workspace CLAUDE.md — the platform-level orientation document
• Final decisions in .claude/docs/ (settled architecture decisions derived from completed brainstorms)
• Wiki + journal in .claude/wiki/ and .claude/journal/ (per the knowledge system)

This is the natural setup for cross-repo work: Claude has the full org as its working set.

Knowledge map

The workspace's CLAUDE.md carries a <!-- wiki:index --> marker block that auto-loads every wiki page's title + summary into Claude's context. See Claude Code conventions for how the marker block works and why it exists.

The wiki itself (.claude/wiki/*.md) is the canonical record of platform-wide patterns, conventions, discoveries, and anti-patterns the maintainer needs handy when working on cross-repo concerns. Pages are kept current — the knowledge system is replace-style for current truth, append-only journal for history.

End of session

When wrapping up:

./scripts/status.sh        # confirm everything is on main + clean
./scripts/push-all.sh      # see what's unpushed

For a more thorough end-of-session pass, /repo-cleanup automates: save-learnings → branch + commit + push + PR + auto-merge → tag + registry + branch prune. Run it from inside Claude Code in the workspace.

Related

• Install the atl CLI — if you only want to USE atl, skip the workspace
• Team-repo maintenance — the discipline every peer-repo PR follows
• Governance — branch protection + the team-repo-maintenance rule pair
• Knowledge system — the journal + wiki layers the workspace's .claude/ directory uses