- Home
- Skills
- Code Quality & Review
- Instruction Enhancer
Instruction Enhancer
Audit your CLAUDE.md and AGENTS.md: cut the lines that don't change agent behaviour, fix the ones the repo contradicts, and add a few proven rules with your real commands.
Free
Works with the AI Agents you already use
Instruction Enhancer
Example session with this skill installed
My CLAUDE.md is getting ignored and feels way too long. Audit my agent instructions and prune the fluff.
- Read your context and instructions
- Compiled the instruction enhancer
I have audited CLAUDE.md and AGENTS.md against your current repository state (using package.json and pnpm-lock.yaml as truth). Your instructions contained 42 lines of directory tours that the agent can already see and several commands that pointed to a deprecated npm workflow despite the repo using pnpm.
Instruction Enhancement Plan
CLAUDE.md — 184 lines → 62 lines
- Remove
- L10-45: Manual directory tour (redundant to file tree).
- L55: "Write clean code" (generic behavioral fluff).
- L60-65: Duplicated test instructions.
- Rewrite
- L8: "Always be thorough" → "Verify changes with
pnpm testbefore reporting completion." - L12: "Use npm for tasks" → Update to
pnpmbased on lockfile.
- L8: "Always be thorough" → "Verify changes with
- Add
- Rule: Search prior art before building new components.
- Don't touch:
dist/,build/, and.next/directories.
- Ask
- L15 mentions a
deploy.shscript that does not exist in the root orscripts/. Should this be removed or updated to the new CI flow?
- L15 mentions a
Next steps
- Review the removal of the directory tour.
- Confirm if the
deploy.shreference is obsolete. - Say "approved" to apply these changes to both
CLAUDE.mdandAGENTS.md.
Connects securely to your tools. The creator never sees your data.
About this skill
Agents read CLAUDE.md and AGENTS.md on every task, so every line costs tokens and attention, and generic or out-of-date lines make agents worse, not better. Instruction Enhancer makes these files shorter and sharper. It removes generic advice, directory tours, and duplicates. It checks every path, command, and version against the repo, using git log to find changes made after the file was last edited. It rewrites vague lines such as "be thorough" into checkable rules. It adds rules only from a fixed library of seven core rules and two UI rules, and only when the repo shows the signal that each rule needs. Commands and do-not-touch paths come from files it reads during the run, never from guesses. When a repo has both CLAUDE.md and AGENTS.md, both get the same content. It shows a remove, rewrite, and add plan, then stops until you approve it.
It checks that documented commands and paths exist and match the files that define them, but it doesn't run the commands. A command that exists but behaves differently than documented can slip through. Behavioural rules come only from its built-in library. It suggests hooks for "always run X" lines but doesn't install them unless asked.
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
Fresh listing
Recently published to Agensi
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 16 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes