CLAUDE.md — [Project Name]
This is the always-loaded context skeleton. Fill every section with real specifics; delete the prompts in brackets. Keep it tight — this is read every session, so every line should earn its place. When something here goes stale, fix it the same day.
Delete this blockquote when you fork.
What this project is
[One paragraph. What it does, who it’s for, what stage it’s at. The thing a new agent needs to not be confused.]
Stack
- Language / runtime: […]
- Framework: […]
- Data: […]
- Deploy target: […]
- Key dependencies and why: […]
How the code is organized
[The map. Where things live, what the top-level directories are for, where NOT to put things. Point at the 3–5 files that matter most.]
Conventions
The idioms an agent must match instead of using its defaults:
- Naming: […]
- Error handling: […]
- Tests: [where, what framework, what “tested” means here]
- Comments: [your bar — almost certainly “why, not what”]
- Imports / structure: […]
Standards — non-negotiable
- Run the no-slop review pass before reporting any non-trivial change done.
- For anything with a surface, follow
DESIGN.md. - Never claim done without proof. “Deployed” / “works” / “passing” requires verification in the same turn — a test run, a curl, a screenshot, a log line.
- Report outcomes faithfully — failing tests get reported as failing, with output.
Git & shipping
- [Branch strategy — e.g. never push to main, feature branches + PRs]
- [How deploys happen, and how to verify one landed]
- [What “done” means — the review gate for this project]
How to work with me
- [Parallelism default — e.g. independent pieces get parallel agents, not a serial chain]
- [Tone / response style you want]
- [What to ask about vs. decide autonomously]
- [Things that have bitten us before — see
memory/]
Pointers
- Playbooks:
playbooks/— procedures for repeated work - Memory:
memory/— durable facts, decisions, scars - Design:
DESIGN.md