ai-orchestration-repo
CLAUDE.md0 starsChanged 3 months agoMITView on GitHub
This repository is an orchestrator. It contains no application code. It holds the documentation for a system whose services live in separate repositories, and the skills that plan, build and review a feature across several of them at once. The services are cloned into projects/, which is gitignored. Each is an autonomous repository with its own .git, CLAUDE.md, skills and CI, and none of them knows this orchestrator exists. Pipeline, all user-invoked: /system-map → /feature-plan → /feature-plan-review → /feature-build (or…
In this repository6 other files it gives its agents.
- feature-buildSkill
- feature-build-step-by-stepSkill
- feature-code-reviewSkill
- feature-plan-reviewSkill
- feature-planSkill
- system-mapSkill
The file
We split this CLAUDE.md into the rules it gives an agent
1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 47
After a feature lands, the seam files of the services it changed are stale by definition.
≈ “Common” is an estimate: at least 1 in 5 of the 27,641 CLAUDE.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
ai-orchestration-repo
This repository is an orchestrator. It contains no application code. It holds the documentation for a system whose services live in separate repositories, and the skills that plan, build and review a feature across several of them at once.
The services are cloned into projects/, which is gitignored. Each is an autonomous
repository with its own .git, CLAUDE.md, skills and CI, and none of them knows this
orchestrator exists.
Layout
| Path | What it is |
repos.yaml | The service registry. Read by setup.sh and by every skill |
setup.sh | Clones or updates the services listed in the registry |
projects/ | The cloned service repositories. Gitignored, never committed |
.claude/rules/principles.md | Always in context: the invariant and how to work across repos |
.claude/rules/definition-of-done.md | Always in context: what "done" means |
.claude/rules/services/<svc>.md | One service's seam. Loads only when that service is touched |
.claude/agents/ | The four working roles; three of them cannot write, by construction |
.claude/lib/build-engine.md | The build procedure, shared by both build skills |
docs/system/ | Architecture, infrastructure, decisions. Read deliberately, never auto-loaded |
docs/features/<slug>/ | One feature: PRD, plan, and its review reports |
Commands
./setup.sh # clone or update every service in repos.yaml
tail -f .claude/logs/instructions.log # what lazily loaded, and whenPipeline, all user-invoked: /system-map → /feature-plan → /feature-plan-review
→ /feature-build (or /feature-build-step-by-step) → /feature-code-review.
Gotchas
gitwith no-Cacts on this repository, not on a service. Every service operation isgit -C projects/<svc> ….- A service's hooks and
settings.jsondo not run here. Claude Code reads those only from the session root. Guardrails live in this repo's.claude/settings.json; real enforcement is server-side branch protection. - New agent files need a restart. Claude Code only watches directories that existed when the session started.
.claude/rules/is not a docs folder. Any.mdthere withoutpaths:frontmatter loads into every session, forever. Do not put a README in it.- After a feature lands, the seam files of the services it changed are stale by
definition. Refresh with
/system-map --only <svc,...>.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

