- Home
- Skills
- Prompt & Skill Engineering
- CLAUDE.md Trim (Keep Every Rule)
Works with the AI Agents you already use
CLAUDE.md Trim (Keep Every Rule)
Bring an oversized CLAUDE.md or AGENTS.md back under Claude Code's size warning without losing a rule.
Free
CLAUDE.md Trim (Keep Every Rule)
Example session with this skill installed
Claude Code says my CLAUDE.md is too large. Trim it without losing anything.
- Read your context and instructions
- Compiled the claude md trim
What I ran
CLAUDE.md: 19,215 chars
130,785 under the limit (target 135,000)
biggest sections:
3,393 ## READ THIS FIRST: what does not exist yet
1,967 ## Gotchas
1,882 ## The schedule and the due rule are shared with the website
1,406 ## Finishing pipeline (new-app skill, step 8)
1,117 ## Architecture
952 ## Commands
929 ## No coin-and-earn
909 ## Analytics
717 ## Languages
703 ### Dependency injection
698 ### BLoC conventions
600 ## What this is
596 ## Scripts
543 ## The look: a house and a timeline
542 ## The store-review prompt
What it means
This file is well under the warning, so there's nothing to trim, and the skill says so instead of cutting anyway. When a file is over, the same section ranking shows where the evidence lives, and that evidence gets moved to docs/ with a link left behind.
Next steps
Nothing for this file. Run it again before adding a big new section, or the moment Claude Code shows its size warning. --links checks that every link still resolves after a trim.
Output is from a real run on one of my projects.
Connects securely to your tools. The creator never sees your data.
About this skill
CLAUDE.md files grow. Each rule picks up the story of why it exists, until Claude Code warns that the file is over its size limit.
This skill keeps every rule in place and moves the evidence behind it (the incident, the measurements, the long example) into docs/, leaving the rule and a link. Then it checks that every link resolves and that no rule was dropped. Numbers, constants and limits stay in the rule itself.
Works for any project and any language. It also handles AGENTS.md, CLAUDE.local.md, imported files and .claude/rules. The default limit matches Claude Code's current 150k-character warning; pass --limit if your version warns at a different size.
Script: check_size.py. Python 3.9+, no packages.
These are free and MIT-licensed, and the source is at github.com/jameskrupnik/indie-app-skills. If you ship Flutter apps, my paid skills cover the release side: new-app, ship-ios, bug-hunt and daily-update.
Only tested in Claude Code. The size limit Claude Code warns at has changed between versions; match --limit to the warning you actually see.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 1 month ago
- Free with an Agensi account
Permissions
File Scopes
Reads and edits CLAUDE.md, AGENTS.md and their imported files, and writes the moved evidence into docs/. Runs python3. No network.