claude-harness
CLAUDE.md79 starsChanged 4 months agoMITView on GitHub
https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
In this repository6 other files it gives its agents.
- checkpointSkill
- flowSkill
- mergeSkill
- prd-breakdownSkill
- setupSkill
- startSkill
The file
We split this CLAUDE.md into the rules it gives an agent
5 rules. About 0 are common, 3 are distinctive, and 2 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.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 26
All GitHub operations use the gh CLI (issues, PRs, merges).
- Very specific: line 30
Always update .claude-harness/claude-progress.json after completing work
- Distinctive: line 31
Update changelog in README.md
- Distinctive: line 32
Commit with descriptive messages
- Distinctive: line 33
Leave codebase in clean, working state
≈ “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
Claude Harness Plugin
Project Overview
Claude Code plugin for automated, context-preserving coding sessions with 5-layer memory architecture, failure prevention, self-improving skills, feature tracking, GitHub integration, Agent Teams with ATDD (Acceptance Test-Driven Development), and quality gate hooks.
Tech Stack
- Shell/Bash (setup.sh, hooks/)
- Markdown (skills)
- JSON (configuration, state files)
Session Startup Protocol
The SessionStart hook auto-injects session context (features, GitHub info, interrupt markers). Before starting harness work, read .claude-harness/session-briefing.md for full context; check .claude-harness/features/active.json for current priorities.
Project Structure
skills/- Agent Skills (SKILL.md with YAML frontmatter, auto-discovered by Claude Code)agents/- Plugin subagents (harness-implementerexecutes delegated feature lifecycles)hooks/- Session hooks (8 registrations, extensionless scripts withrun-hook.cmdpolyglot wrapper)scripts/- Shared helpers (compile-briefing.pycompiles memory into briefings)schemas/- Canonical JSON Schema files for state file validationsetup.sh- Project initialization script (memory dirs, CLAUDE.md, migrations; stamp-gated at session start).claude-plugin/plugin.json- Plugin manifest- Marketplace lives in separate repo:
panayiotism/claude-harness-marketplace - Updates tracked by git commit SHA (no semver in plugin.json) — GitHub Actions syncs to marketplace on every push
GitHub Integration
All GitHub operations use the gh CLI (issues, PRs, merges). The GitHub MCP server is NOT required.
Development Rules
- Work on ONE feature at a time
- Always update
.claude-harness/claude-progress.jsonafter completing work - Update changelog in
README.md - Commit with descriptive messages
- Leave codebase in clean, working state
Available Skills (6 total)
/claude-harness:setup- Initialize harness in project/claude-harness:start- Start session, compile context/claude-harness:flow- Unified workflow (recommended)- Flags:
--no-merge--plan-only--autonomous--quick--fix--team
- Flags:
/claude-harness:checkpoint- Manual commit + push + PR/claude-harness:merge- Merge all PRs, auto-version, release/claude-harness:prd-breakdown- Analyze PRD and break down into features with Gherkin acceptance criteria
Schema Versioning Convention
- The
"version"field in JSON state files refers to the data schema version, not the plugin version - Increment schema version only when the shape of the data changes in a breaking way
- Canonical schemas live in
schemas/*.schema.json— reference these, don't embed inline examples
Progress Tracking
See: .claude-harness/claude-progress.json and .claude-harness/features/active.json
https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
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.

