Skip to main content
skill CX Codex

chisle-audit — Claude Code skill: audit and install

The chisle-audit Claude Code skill checks code and prose for excess. Run it on a file, a diff, or a repository to get a ranked list of potential cuts.

MIT ★ 1,250 updated 2026-11-14

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add JayPokale/Chisle --skill chisle-audit -a claude-code
Open on GitHub

What chisle-audit does

chisle-audit looks for overbuilt code and wordy documentation. It can flag hand-written substitutes for standard library features, abstractions with a single implementation, and settings that never change. It also spots repeated explanations, stale sections, and comments that merely describe the code.

With no argument, it audits the current git diff; if that is empty, it checks HEAD~1..HEAD. A path targets a file or directory. A whole-repository request walks the tree while skipping vendored or generated files, node_modules, dist, and lockfiles. Findings are ranked by estimated savings, with uncertain ones marked. The skill is read-only: it does not edit files or apply fixes.

What it helps with

  • spot a homegrown version of a standard library feature
  • trim comments that repeat what the code does
  • find abstractions and settings with no current need
  • locate stale or padded documentation

Example prompts for the agent

  • Audit this file for overbuilt code and wordy prose.
  • Review the current diff and rank the biggest possible cuts.
  • Scan the whole repository for code and documentation bloat.

How to install chisle-audit

1Skills CLI (Vercel)

npx skills add JayPokale/Chisle --skill chisle-audit -a claude-code

2Manually: put the files in place

~/.claude/skills/chisle-audit/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/chisle-audit/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about chisle-audit

How do I install chisle-audit in Claude Code?

Run npx skills add JayPokale/Chisle --skill chisle-audit -a claude-code. The skill file is at ~/.claude/skills/chisle-audit/SKILL.md.

What does chisle-audit do?

It checks code and prose for excess, then returns a ranked report of potential cuts. It does not change files.

What does chisle-audit inspect when called without an argument?

It checks the current git diff, including staged and unstaged changes. If the diff is empty, it audits HEAD~1..HEAD.

Source text

SKILL.md · MIT
---
name: chisle-audit
description: >
  One-shot efficiency audit of a file, diff, or whole repo across BOTH axes at
  once: over-engineered code (reinvented stdlib, needless abstractions,
  speculative config) AND bloated prose (verbose comments, padded docstrings,
  redundant doc sections). Neither a pure code-minimizer nor a pure prose
  compressor does both in one pass. That's the point. Ranked report, biggest
  saving first; changes nothing. Use when the user says "chisle audit", "/chisle-audit",
  "audit this for bloat", "what can I cut", "review this PR for over-engineering
  and verbosity".
---

# Chisle Audit

Scan the target (a diff, a file, or the repo tree) and report what to cut, on
both axes. One-shot. Read-only: never edit, never write a flag, never apply fixes.

## Scope

- No argument → audit the current `git diff` (staged + unstaged). Empty diff → audit `HEAD~1..HEAD`.
- A path → audit that file or directory.
- "repo" / "whole repo" → walk the tree (skip vendored/generated/`node_modules`/`dist`/lockfiles).

## What to flag

**Code (the YAGNI axis):**
- Reinvented stdlib (hand-rolled debounce, deep-clone, groupBy, retry loop, date math)
- Abstraction with one implementation (interface/factory/wrapper for a single case)
- New dependency for what a few lines or an installed dep already covers
- Config/option/flag that never varies
- Speculative "for later" scaffolding with no current caller
- Verbose code where a native platform feature (CSS, DB constraint, `<input type>`) does it

**Prose (the compression axis), the half a code-only auditor misses:**
- Comments that restate the code (`i += 1  // increment i`)
- Docstrings/READMEs padded with filler, hedging, ceremony, or duplicated content
- Multi-paragraph explanations where one tight sentence carries the meaning
- Decorative tables/emoji/headings that add tokens, not information
- Dead prose: TODO graveyards, stale "see also" links, obsolete sections

## What NOT to flag

Input validation at trust boundaries, error handling that prevents data loss,
security, accessibility, deliberate `// chisle:` / `// ponytail:` shortcuts already
documented, or domain comments that explain *why* (not *what*).

## Output

One ranked list, biggest cut first. One finding per line. No preamble, no praise.

```
path:line  [code|prose]  <what's bloated> → <the lean replacement>. (~N lines/tokens)
```

End with a two-line summary:

```
N findings: X code, Y prose. Est. removable: ~A lines code, ~B lines prose.
Biggest win: <the single highest-impact cut>.
```

Be honest about uncertainty: mark a finding `(check)` if cutting it might lose
behavior you can't verify from the snippet. Lean toward fewer, high-confidence
findings over a long speculative list.

© JayPokale · MIT · JayPokale/Chisle

Similar skills

SKILL.md skills catalog →

From the same set4

Project sponsors

Companies that keep this analytics open