agentille-claude-md
Skill1 starsChanged 20 days agoMITView on GitHub
Tune a CLAUDE.md to be lean and high-signal. Applies a fixed "less is more" rubric and proposes a shorter rewrite with a per-line cut-list. Non-destructive: shows a diff, backs up the original, writes only on explicit approval. Defaults to ~/.claude/CLAUDE.md; pass a path for a project file. Runs only when the user types /agentille-claude-md; model invocation is disabled.
In this repository5 other files it gives its agents.
- .claude/CLAUDE.mdCLAUDE.md
- agentille-initSkill
- agentille-projectSkill
- agtSkill
- stormSkill
The file
We split this skill into the rules it gives an agent
9 rules. About 0 are common, 4 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Target: No argument → ~/.claude/CLAUDE.md.
- Very specific: line 18
On no: write nothing and print No changes written.
- Distinctive: line 22
Never write without explicit approval.
- Very specific: line 23
Always back up first; never overwrite a .bak unasked.
- Distinctive: line 24
Never create a file; if the target is absent, stop.
- Distinctive: line 25
Never anonymize or strip the user's name, role, stack or genuine constraints.
- Distinctive: line 26
Never add a rule the original did not state: a rewrite merges and shortens, it does not invent.
- Very specific: line 30
Before (4 lines): I am a developer and would like you to be concise.
- Very specific: line 32
After (2 lines): - Developer.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill 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
agentille-claude-md — CLAUDE.md tune-up
Opinionated trim, never a silent overwrite. Runs locally: reads one file, writes that file plus a .bak, sends nothing anywhere. Runs only when typed (disable-model-invocation).
Steps
- Target. No argument →
~/.claude/CLAUDE.md; otherwise the given path. If the file does not exist, say "No CLAUDE.md at<target>— nothing to tune" and STOP; create nothing. Otherwise read it and count its lines. - Rewrite in memory. Apply
rubric.mdto produce a leaner REWRITE. Identity and personal context (rubric point 7) stay verbatim. Tag every removed or merged line with a reason tag from the rubric. - Present: the full REWRITE, the cut-list (one line per removed or merged original line with its tag), and
before <N> → after <M>. Then ask: "Apply this rewrite to<target>? (yes / no)". - On yes:
cp <target> <target>.bak(if the.bakexists, ask before overwriting it; on no, stop without writing), then write REWRITE to<target>. PrintTuned <target> — N → M lines. Backup: <target>.bak. On no: write nothing and printNo changes written.
Hard rules
- Never write without explicit approval; diff first.
- Always back up first; never overwrite a
.bakunasked. - Never create a file; if the target is absent, stop.
- Never anonymize or strip the user's name, role, stack or genuine constraints.
- Never add a rule the original did not state: a rewrite merges and shortens, it does not invent.
Example (generic)
Before (4 lines): I am a developer and would like you to be concise. / Use TypeScript for this Vite app. / Be helpful and write good clean code. / Use TypeScript.
After (2 lines): - Developer; be concise. / - TypeScript.
Cut-list: "Vite app" → inferable; "Be helpful…" → vague; "Use TypeScript." → duplicate.
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.

