ekstremediaekstremedia/claude-code-plan-and-execution/docs/llms.txt

claude-code-plan-and-execution

llms.txt1 starsChanged 25 days agoMITView on GitHub

A two-command Claude Code workflow: /make-plan writes a self-contained plan file with an expensive model, /execute-plan runs it in a fresh session with a cheap orchestrator that delegates every code change to model-pinned worker agents and gates the risky phases behind an expensive read-only reviewer. The links below point at raw markdown — that is the complete content, no rendering needed. The skill and agent files are the workflow itself, not documentation about it.

In this repository3 other files it gives its agents.
- AGENTS.mdAGENTS.md
- execute-planSkill
- make-planSkill

The file

Create a free account to mount

Files

llms.txt

# claude-code-plan-and-execution

> A two-command Claude Code workflow: /make-plan writes a self-contained plan
> file with an expensive model, /execute-plan runs it in a fresh session with a
> cheap orchestrator that delegates every code change to model-pinned worker
> agents and gates the risky phases behind an expensive read-only reviewer.

The links below point at raw markdown — that is the complete content, no
rendering needed. The skill and agent files are the workflow itself, not
documentation about it.

## Docs

- [Overview](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/docs/index.md): What the workflow is, the shape in one screen, and the scope of the claims.
- [Design](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/docs/design.md): The six roles, the phase packet, the git discipline that closes a real review hole, and when not to use any of this.
- [Gotchas](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/docs/gotchas.md): The failure modes that break the setup silently. `scripts/doctor.sh` asserts the checkable ones.
- [Prompting](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/docs/prompting.md): What changed with Opus 5 and Sonnet 5, and the prompt consequences — one inverts standard advice.
- [The advisor](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/docs/advisor.md): What /advisor is, what is verified about it, and why it is off during orchestrated execution.
- [Composing](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/docs/composing.md): Working alongside skill packs this repository does not depend on — key on the artifacts, keep the references soft, keep one reviewer per change.
- [README](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/README.md): Install, use, verification, and the four things that break it.

## The workflow files themselves

- [make-plan skill](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/skills/make-plan/SKILL.md): The planner. Research budget, plan-file contract, handoff.
- [execute-plan skill](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/skills/execute-plan/SKILL.md): The orchestrator loop. Delegation, review, gates, integration.
- [planning-researcher agent](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/agents/planning-researcher.md): Read-only breadth search during planning.
- [implementer agent](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/agents/implementer.md): Writes the code and the tests, one phase at a time.
- [quick-implementer agent](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/agents/quick-implementer.md): Mechanical, exactly-specified work. Refuses judgment.
- [plan-reviewer agent](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/agents/plan-reviewer.md): Read-only risk gate over the real changes.
- [Plan template](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/templates/PLAN-TEMPLATE.md): The plan-file contract every plan fills in.

## Optional

- [Test wrapper example](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/templates/bin/test-example.sh): The wrapper convention that keeps test logs out of agent context.
- [Settings snippet](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/templates/settings.snippet.json): plansDirectory and the read-only permission allowlist.
- [CLAUDE.md snippet](https://raw.githubusercontent.com/ekstremedia/claude-code-plan-and-execution/main/templates/CLAUDE.md.snippet.md): The two-line routing note, and why it is deliberately that short.

Download now (ZIP)

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free