jf-agentic-coding-starter-kit
CLAUDE.md1 starsChanged 4 months agoMITView on GitHub
Technology-neutral, repository-local workflow for Claude Code and Codex. The project stack is chosen by /3-csk-architecture, never by this template. 1. Inspect the skills and tools available in the runtime and this repository. 2. Read every applicable skill before acting; explicitly requested skills are mandatory. Prefer a suitable skill over an improvised workflow. 3. Run /csk-start before product or implementation work. It validates .csk/project-state.json, reads durable work from tasks/INDEX.md, and reconciles live Git state. 4. On the first project use only,…
In this repository18 other files it gives its agents.
- AGENTS.mdAGENTS.md
- 1-csk-initSkill
- 2-csk-write-specSkill
- 3-csk-architectureSkill
- 4-csk-frontendSkill
- 5-csk-backendSkill
- 6-csk-qaSkill
- 7-csk-deploySkill
- audit-plan-loopSkill
- claude-skill-proxy-syncSkill
- create-worker-worktreeSkill
- csk-adopt-plan-scaffoldSkill
- csk-helpSkill
- csk-refineSkill
- csk-startSkill
- finish-branchSkill
- pull-main-ffSkill
- review-loopSkill
The file
We split this CLAUDE.md into the rules it gives an agent
11 rules. About 0 are common, 5 are distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 9
Read every applicable skill before acting.
- Very specific: line 11
Run /csk-start before product or implementation work.
- Very specific: line 18
Follow .claude/rules/workflow-state.md for the one authoritative routing and status-transition model.
- Distinctive: line 21
Use deterministic tools for repeatable mechanical work such as parsing, validation, locking, and atomic state changes.
- Distinctive: line 31
Never install or synchronize kit skills into a user's global Claude or Codex directories.
- Very specific: line 33
Run /claude-skill-proxy-sync after adding, renaming, moving, or removing a canonical skill or changing its trigger/interface metadata.
- Very specific: line 47
Use csk.config.json only for repository-local path/workflow adaptation.
- Distinctive: line 48
Do not select frameworks, runtimes, package managers, storage, tests, release targets, or generated scaffolds before architecture approves …
- Distinctive: line 51
Create a task only for work that remains open/blocked, spans a session, is deferred, or is a verified interruption point.
- Very specific: line 53
Use /audit-plan-loop for non-trivial plans and /review-loop for productive artifacts according to .claude/rules/loop-policy.md.
- Very specific: line 58
After verified work on a non-default branch, route to /finish-branch for safe commit, integration, and optional cleanup.
≈ “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 Coding Starter Kit
Technology-neutral, repository-local workflow for Claude Code and Codex. The
project stack is chosen by /3-csk-architecture, never by this template.
Start every repository session
- Inspect the skills and tools available in the runtime and this repository.
- Read every applicable skill before acting; explicitly requested skills are mandatory. Prefer a suitable skill over an improvised workflow.
- Run
/csk-startbefore product or implementation work. It validates.csk/project-state.json, reads durable work fromtasks/INDEX.md, and reconciles live Git state. - On the first project use only,
/csk-startinventories the repository and asks whether existing code or product/planning evidence must be adopted. The tracked answer applies to all clones after its commit reaches the default branch.yesroutes to/csk-adopt-plan-scaffold;nocontinues greenfield. - Follow
.claude/rules/workflow-state.mdfor the one authoritative routing and status-transition model.
Use deterministic tools for repeatable mechanical work such as parsing, validation, locking, and atomic state changes. Use LLM context judgment for semantic relevance, ambiguity, design tradeoffs, and decisions that cannot be reduced to one fixed algorithm. The opposite of deterministic is non-deterministic.
Local skills and agent compatibility
- Canonical skills live in
.claude/skills/; ready Codex proxies live in.codex/skills/.AGENTS.mdis the Codex entry point. - Never install or synchronize kit skills into a user's global Claude or Codex directories.
- Run
/claude-skill-proxy-syncafter adding, renaming, moving, or removing a canonical skill or changing its trigger/interface metadata. Users can use the same skill to sync their own repository-local additions. - Required workflow skills include numbered steps 1–7,
/csk-start,/csk-adopt-plan-scaffold,/finish-branch,/create-worker-worktree,/pull-main-ff,/audit-plan-loop,/review-loop,/csk-refine, and/csk-help.
Core workflow rules
- Sources of truth: product in
docs/PRD.md; whole-product contract indocs/master-feature.md; quality indocs/engineering-principles.md; technology indocs/architecture.md; feature state/details infeatures/; durable open work intasks/INDEX.md. - Use
csk.config.jsononly for repository-local path/workflow adaptation. - Do not select frameworks, runtimes, package managers, storage, tests, release targets, or generated scaffolds before architecture approves them.
- Preserve unrelated user changes; read before editing and verify after writing.
- Create a task only for work that remains open/blocked, spans a session, is deferred, or is a verified interruption point. Do not create same-turn churn.
- Use
/audit-plan-loopfor non-trivial plans and/review-loopfor productive artifacts according to.claude/rules/loop-policy.md. - QA findings return to the owning implementation skill and remain
In Reviewuntil verified. Releases require both a pre-release readiness review and a final review of the actual result beforeDeployedbookkeeping. - After verified work on a non-default branch, route to
/finish-branchfor safe commit, integration, and optional cleanup. - Verification never authorizes commit, push, merge, pull request, tag, release, deployment, or deletion. Obtain explicit authority for those actions.
Workflow
/1-csk-init → /2-csk-write-spec → /3-csk-architecture →
/4-csk-frontend and/or /5-csk-backend → /6-csk-qa →
/finish-branch when needed → /7-csk-deploy
@.claude/rules/general.md @.claude/rules/workflow-state.md @.claude/rules/loop-policy.md @.claude/rules/adapter-config.md @docs/PRD.md @docs/master-feature.md @docs/engineering-principles.md @docs/architecture.md @features/INDEX.md @tasks/INDEX.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.

