plan-critique-skills
CLAUDE.md13 starsChanged 2 months agoMITView on GitHub
This project provides five planning skills for three agents. - Claude Code reads them from skills/ as create, critique, cycle, execute and archive. Installed as a plugin they are invoked as /plan:create, /plan:critique, /plan:cycle, /plan:execute and /plan:archive; installed manually they are invoked bare as /create, /critique, etc. - Antigravity reads them from .gemini/skills/ (registered via .agents/skills.json or ~/.gemini/config/skills.json) as plan-create, plan-critique, plan-cycle, plan-execute and plan-archive. - GitHub Copilot CLI reads them from .github/skills/ as plan-create, plan-critique, plan-cycle, plan-execute and plan-archive,…
In this repository10 other files it gives its agents.
- plan-archiveSkill
- plan-createSkill
- plan-critiqueSkill
- plan-cycleSkill
- plan-executeSkill
- archiveSkill
- createSkill
- critiqueSkill
- cycleSkill
- executeSkill
The file
We split this CLAUDE.md into the rules it gives an agent
3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 53
Never use emojis in the markdown text.
- Distinctive: line 58
Follow semver: MAJOR for breaking changes, MINOR for new features, PATCH for fixes.
- Distinctive: line 59
Every release must update six things:
≈ “Common” is an estimate: at least 1 in 5 of the 26,980 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
Project Standards
This project provides five planning skills for three agents.
- Claude Code reads them from
skills/ascreate,critique,cycle,executeandarchive. Installed as a plugin they are invoked as/plan:create,/plan:critique,/plan:cycle,/plan:executeand/plan:archive; installed manually they are invoked bare as/create,/critique, etc. - Antigravity reads them from
.gemini/skills/(registered via.agents/skills.jsonor~/.gemini/config/skills.json) asplan-create,plan-critique,plan-cycle,plan-executeandplan-archive. - GitHub Copilot CLI reads them from
.github/skills/asplan-create,plan-critique,plan-cycle,plan-executeandplan-archive, invoked as/plan-create,/plan-critique,/plan-cycle,/plan-executeand/plan-archive.
These skills help create, review, execute, and archive plans written by the user.
The cycle skill is a shortcut through create and critique: it drafts plan.md only when the user has not
written one, then critiques and merges up to cycleIterations times (default 3) and stops for approval. It
merges on its own only the changes that are obvious, and asks the user about anything ambiguous, anything that
touches what the user wrote, and anything that changes the scope of a chapter. It never executes the plan.
Architecture
Session tracking for parallel plans:
- Each session tracks its current plan via a file in
.planning/.sessions/. - Claude Code and Antigravity key the file by the process ID obtained via
echo $PPID, and clean up stale files opportunistically viakill -0 [PID]checks. - Copilot CLI keys the file by the
COPILOT_AGENT_SESSION_IDenvironment variable, and cleans up session files that have not been modified in the last 7 days. It never shells out to$PPIDorkill, because those do not work on Windows. - Skills read the session file to determine the current plan and show it as "(current session)" in selection.
- The
.sessions/directory is local state and should be gitignored.
Per agent settings, manifests, and documentation:
- Claude Code:
docs/claude-code.md,.claude-plugin/plugin.json,.claude-plugin/marketplace.json, config in.claude/plan-critique-config.json, project standards inCLAUDE.md. - Agy / Antigravity:
docs/agy.md, config in.gemini/plan-critique-config.jsonwith a fallback read of.claude/plan-critique-config.json, project standards inGEMINI.md,AGENTS.mdorCLAUDE.md. - Copilot CLI:
docs/copilot.md,.github/plugin/plugin.json,.github/plugin/marketplace.json, config in.copilot/plan-critique-config.jsonwith a fallback read of.claude/plan-critique-config.json, project standards in.github/copilot-instructions.md,AGENTS.mdorCLAUDE.md. - The Copilot skills carry
name,descriptionandlicensefrontmatter. They do not carryallowed-tools,argument-hint,modeloreffort, which are Claude Code fields.
Code Style
- The skill markdown files in the
skills/,.gemini/skills/and.github/skills/folders (eachSKILL.mdand its supporting.mdfiles) should have a word wrap of maximum 120 characters. - The markdown text should be minimal, do not use bold or italic text unless absolutely necessary.
- The markdown text should be mostly plain text paragraphs, ordered lists, emphasys on code with backticks and some code blocks.
- The markdown text should include thematic breaks (
---) to separate sections. - The text from the skill markdown files should be written in plain English and easy to read from a text editor even if markdown support is not available.
- Never use emojis in the markdown text.
Versioning and Releases
- The canonical version lives in
.claude-plugin/plugin.jsonunder theversionfield. - Follow semver: MAJOR for breaking changes, MINOR for new features, PATCH for fixes.
- Every release must update six things:
versionfield in.claude-plugin/plugin.jsonversionfields in.github/plugin/plugin.jsonand.github/plugin/marketplace.json- The hardcoded version in the banners inside
skills/create/SKILL.mdandskills/cycle/SKILL.md - The same banners inside
.gemini/skills/plan-create/SKILL.md,.gemini/skills/plan-cycle/SKILL.md,.github/skills/plan-create/SKILL.mdand.github/skills/plan-cycle/SKILL.md CHANGELOG.mdwith a new section describing the changes- A git tag in the format
v[VERSION](e.g.v1.0.0)
Documentation
- Short and succinct
- The README should have three major sections:
- Install
- Marketplace installation (recommended), per agent
- Manual installation via git clone, per agent
- Usage
- Pre-requisites
- Skills overview
- Workflow steps
- Credits
- Install
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.

