agentleFS
schen22schen22/claude-config/CLAUDE.md

claude-config

CLAUDE.md0 starsChanged 2 days agoMITView on GitHub

A project's own CLAUDE.md wins where it is more specific. - Plan, then wait. Before changing files: a plan table (| Step | Files | Lane | Technique | Notes |; Lane names who runs it and which steps can run side by side because they share no files) plus open questions, and what you think and why. Execute after Sarah confirms. - Verify, then argue. Check current behaviour (build, run, read) before describing it. Before proposing a claim or…

The file

We split this CLAUDE.md into the rules it gives an agent

7 rules. About 0 are common, 3 are distinctive, and 4 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.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Plan, then wait: Before changing files: a plan table (| Step | Files | Lane | Technique | Notes |.

    line 7
  2. Distinctive:

    Verify, then argue: Check current behaviour (build, run, read) before describing it.

    line 11
  3. Distinctive:

    Hyperlink doc references: Every mention of another doc, in files and in replies, is a Markdown link (to the section anchor when there is …

    line 15
  4. Very specific:

    Default is inline, except a request independent of the main thread: that goes to a background helper (delegate) and never supersedes a …

    line 43
  5. Very specific:

    Never two sessions editing one working tree: helpers work in their own worktree, and while Sarah has uncommitted edits so do you (git …

    line 46
  6. Distinctive:

    Before the first edit, run git status.

    line 48
  7. Very specific:

    Every project exposes a fast check for iterating and a full check before commit, each printing ~5 lines.

    line 52

≈ “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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
CLAUDE.md

Every project — Sarah's defaults

A project's own CLAUDE.md wins where it is more specific.

How we work

  • Plan, then wait. Before changing files: a plan table (| Step | Files | Lane | Technique | Notes |; Lane names who runs it and which steps can run side by side because they share no files) plus open questions, and what you think and why. Execute after Sarah confirms.
  • Verify, then argue. Check current behaviour (build, run, read) before describing it. Before proposing a claim or a number, state what would refute it and what its arithmetic assumes. Challenge Sarah's assumptions when the evidence disagrees.
  • Terse. Tables, diffs, short bullets.
  • Hyperlink doc references. Every mention of another doc, in files and in replies, is a Markdown link (to the section anchor when there is one). Never a bare ID: write the name first, the ID in parentheses, linked to its definition, e.g. Repeated context (C2).
  • YAGNI. Build only what the project's size and stakes (working-model) and a named current need justify. Suggest scope changes; never make them unasked.

State

  • One status file per project: PLAN.md — Now / Next / Later / Done, plus a dated, append-only Decisions log. Lessons live in the project CLAUDE.md under Lessons.
  • New projects: scaffold this with the working-model skill. Existing projects keep their scheme unless Sarah asks to migrate.
  • Local auto memory is keyed to one folder and never travels. Don't put workflow lessons there.
  • The global setup (skills, hooks, plugins, decisions, who runs what) is tracked in ~/.claude/PLAN.md. Read it before changing anything in ~/.claude; update it after.

Where a lesson goes (strongest first)

  1. A committed script or test — works without anyone remembering.
  2. A template that must be filled in (e.g. the delegate brief).
  3. A step in a skill.
  4. A line in a CLAUDE.md.

Project-specific → that project's Lessons; applies anywhere → this file or a skill. /end sorts them.

Subagents

  • Default is inline, except a request independent of the main thread: that goes to a background helper (delegate) and never supersedes a running thread unless Sarah says so.
  • Load the delegate skill before any Agent call that starts an implementation or review helper.
  • Never two sessions editing one working tree: helpers work in their own worktree, and while Sarah has uncommitted edits so do you (git worktree add --detach <scratch> HEAD, then cherry-pick).
  • Before the first edit, run git status; the status shown at session start can be out of date.

Gates

  • Every project exposes a fast check for iterating and a full check before commit, each printing ~5 lines. Convention: npm run gate:fast / npm run gate (or the stack's equivalent). If a project lacks them, propose adding them before delegating.

Sharing and privacy

  • A force-push leaves old commits reachable on GitHub by SHA. To purge personal data from a new repo, delete and recreate it (gh auth refresh -s delete_repo), then check the old SHA 404s.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.