di2-starter-kit
CLAUDE.md0 starsChanged 4 months agoMITView on GitHub
{{ONELINEDESCRIPTION}} {{TECH_STACK}} The application source layout (src/, app/, cmd/, …) depends on the chosen stack and is created by /architecture + /backend + /frontend against .claude/rules/stack.md. The workflow is the spine of this template and is the same regardless of tech stack. Steps that only make sense for some stacks (UX, Frontend) are optional and are pruned by /init for projects without a UI. 0. /init - One-time bootstrap. Asks for the product vision + framework conditions (language/runtime, UI yes/no,…
We split this CLAUDE.md into the rules it gives an agent
5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 8
Run /init once: it asks for the product vision + tech-stack framework conditions, fills these placeholders, writes .claude/rules/stack.md, …
- Very specific: line 72
/bug: - Document a bug as docs/bugs/bug-NNNN-<slug>.md and run the bug-loop. - Very specific: line 89
All features are tracked in features/INDEX.md.
- Distinctive: line 104
Human-in-the-loop: every workflow phase has a user approval checkpoint.
- Very specific: line 117
Cookie banner: if the project ships a public web UI, check whether a cookie banner is required (GDPR/ePrivacy) — see …
≈ “Common” is an estimate: at least 1 in 5 of the 25,680 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_NAME}}
{{ONE_LINE_DESCRIPTION}}
<!--
This is a TEMPLATE file. On a fresh project, the `{{PLACEHOLDER}}` values below are
still unfilled — that is the signal that the project has not been initialized yet.
Run `/init` once: it asks for the product vision + tech-stack framework conditions,
fills these placeholders, writes `.claude/rules/stack.md`, and prunes the skills/rules
that this project does not need. See TEMPLATE.md for what `/init` does.
-->Tech Stack
{{TECH_STACK}}
<!--
Until `/init` runs, the concrete stack is undefined. The authoritative, machine-readable
stack lives in `.claude/rules/stack.md` (auto-loaded into every session). Skills that need
to know "how do we build / test / deploy here?" read that file — they do NOT hard-code a stack.
-->Project Structure
.claude/
rules/ Always-on conventions (auto-loaded each session)
skills/ Workflow + cross-cutting skills (slash commands)
.github/
workflows/ CI + DB deploy workflows (manual dispatch; pruned by /init when not needed)
features/
INDEX.md Feature status overview (single source of truth for what's planned/built)
archive/ Completed features (Deployed/Superseded), git mv'd per quarter (<YYYY-QN>/)
docs/
PRD.md Product Requirements Document
bugs/ One file per bug (docs/bugs/bug-NNNN-<slug>.md) + INDEX.md + archive/
kb/ Knowledge-base / troubleshooting articles (kb-NNN-*.md)
setup/ One-time setup guides (setup-NNN-*.md)
production/ Generic production hardening guides
db/ SQL DB artifacts (PostgreSQL): config/ · database/ · schemas/<schema>/ · scripts/ · tests/
(worked example slice; pruned by /init when database == none; tests/ is a
separate opt-out at /init even when a database is kept)
scripts/ Helper scripts (e.g. spin up a new project from this template)The application source layout (src/, app/, cmd/, …) depends on the chosen stack and is
created by /architecture + /backend + /frontend against .claude/rules/stack.md.
Development Workflow
The workflow is the spine of this template and is the same regardless of tech stack. Steps
that only make sense for some stacks (UX, Frontend) are optional and are pruned by /init for
projects without a UI.
/init- One-time bootstrap. Asks for the product vision + framework conditions (language/runtime, UI yes/no, backend yes/no, database, auth, hosting/deploy, CI), writes.claude/rules/stack.md+ this Tech-Stack section, and removes the skills/rules the project does not need. Hands off to/requirements./requirements- Create feature specs from an idea (and the PRD on first run)./architecture- Design the technical approach (PM-friendly, no code). Decides backend yes/no./ux- UX critique & mockups (only if the project has a UI)./backend- Build APIs, data model, server-side logic (only if the feature needs a backend)./frontend- Build the UI against the real APIs (only if the project has a UI)./qa- Test the feature against its acceptance criteria + feature-scoped security checks./review- Code review of the diff against spec & conventions (before the first deploy)./check-updates- Cross-cutting maintenance check (dependencies, base images, CI). Periodic./security- Project-wide security audit. Mandatory gate before/deployto production./deploy <env>- Roll out to an environment (only if the project has a deploy target).
Cross-Cutting Skills (no fixed workflow slot)
/bug- Document a bug asdocs/bugs/bug-NNNN-<slug>.mdand run the bug-loop. Triggered anytime (during/qa, after/deploy, on a user report)./auth- Diagnose & fix authentication problems (only if the project has auth). Generic OIDC / session / login diagnosis (diagnose-first: logs → tokens → state → hypothesis → fix)./help- Context-aware guide: where am I in the workflow, what to do next.
Bug-Loop (parallel to the workflow)
/bug <description>documents it asBUG-NNNNindocs/bugs/bug-NNNN-<slug>.md(statusOpen).- Fix it via
/frontend BUG-NNNN(UI),/backend BUG-NNNN(API/DB/server), or/auth BUG-NNNN(login/session/OIDC) — commit convention in.claude/rules/general.md. /qare-tests →/bug close BUG-NNNNcloses it andgit mvs it into the quarterly archive.
A feature only moves on toward /review → /deploy once it has no open Critical/High bugs.
Feature Tracking
All features are tracked in features/INDEX.md. Every skill reads it at start and updates it when
done. Feature specs live in features/<prefix>-XXXX-name.md.
- Feature-ID scheme:
{{FEATURE_PREFIX}}-XXXX(sequential, zero-padded). The prefix is set by/init; the default in this template isfeat. Skills refer to "the feature-ID scheme defined here" rather than hard-coding a prefix — examples in skill text usefeat-XXXXillustratively. - Commits:
feat({{FEATURE_PREFIX}}-XXXX): description,fix({{FEATURE_PREFIX}}-XXXX): BUG-NNNN; description. - Archive on completion:
Deployed(prod) andSupersededare terminal — the spec isgit mv'd tofeatures/archive/<YYYY-QN>/and itsINDEX.mdrow moves to the "Features (archived)" table (same quarterly scheme as the bug archive). Details in.claude/rules/general.mdandfeatures/README.md.
Key Conventions
- One feature per spec file (Single Responsibility).
- Human-in-the-loop: every workflow phase has a user approval checkpoint.
- Conventions live in
.claude/rules/and auto-load each session. The authoritative tech stack is.claude/rules/stack.md— when a skill needs a stack fact, it reads that file, never guesses. README.mdfiles always start with a Table of Contents.
Build & Test Commands
{{BUILD_TEST_COMMANDS}}
<!-- e.g. for Python: ruff check . · pytest · docker build . — filled by /init from stack.md -->Legal
- Cookie banner: if the project ships a public web UI, check whether a cookie banner is required
(GDPR/ePrivacy) — see
.claude/rules/cookies.md. Pruned by/initfor non-web projects.
Product Context
@docs/PRD.md
Feature Overview
@features/INDEX.md
In this repository14 other files it gives its agents.
- architectureSkill
- authSkill
- backendSkill
- bugSkill
- check-updatesSkill
- deploySkill
- frontendSkill
- helpSkill
- initSkill
- qaSkill
- requirementsSkill
- reviewSkill
- securitySkill
- uxSkill
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
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.

