How to use the Code Comment Cleanup skill (antislop-code)

What the antislop-code skill makes the agent do

When an AI coding agent writes or edits comments, this skill tells it to filter out comments that read as generically AI-written. Its rules cover decorative separators, restated obvious code, workflow narration, empty labels, vague placeholders, signature echoes, decorative emoji and end markers. It also covers over-long comments and stiff or loud wording.

The skill also tells the agent what to keep. Comments explaining business logic, architectural decisions, security considerations, performance trade-offs, concurrency behaviour, protocol details, API contracts, workarounds, edge cases, assumptions and licensing notices should stay. The skill is restricted to comments: it is told never to modify executable code, identifiers, imports, formatting, indentation, whitespace, control flow or logic. When in doubt, it leaves the code alone.

• Removes banner comments built from repeated characters or ALL CAPS labels
• Removes comments that repeat the next line, such as a comment saying Initialize the variable above a variable declaration
• Removes step-numbered narration and generic labels like Main logic
• Shortens over-explained comments to one or two lines
• Keeps comments that explain why, not what

Install the antislop-code skill

Download the ZIP of the skill from this page and unzip it. Place the skill folder in one of two locations. For a personal setup, use ~/.claude/skills/. For a single project, use the .claude/skills/ folder inside that project.

If you use the Claude apps instead, open Settings, go to Capabilities, then Skills, and upload the skill there.

• Personal install: ~/.claude/skills/
• Project install: .claude/skills/ inside your project folder
• Claude apps: Settings → Capabilities → Skills, then upload

Add the companion antislop.md core file

The SKILL.md file refers to a companion core file called antislop.md, and it relies on a Delivery Gate defined in that core. The core file is not included in the files this skill provides. The skill says to load it together with antislop.md whenever a task touches code comments.

Without antislop.md, the skill's rules still describe the comment patterns, but the core mechanism (the purpose test, the three tiers and the Delivery Gate) is missing. Obtain the core file and install it alongside this skill before relying on the results.

Trigger the skill with natural requests

You do not need a special command. The skill is meant to load when the task writes or edits code comments, so describe that task in ordinary words. Name the file or code you want cleaned, and say what you want from the comments.

Requests that match the skill's purpose include asking the agent to tidy the comments in a file, remove filler comments from code it just generated, or review its comments before you commit. Say explicitly if you want useful comments kept as they are.

• Clean up the comments in src/checkout.js, keep anything about business rules or security
• Remove decorative banners and restating comments from the auth module
• Review the comments the agent just wrote and cut the filler, but do not change any code
• Shorten the over-long comments in this file to one or two lines each

Example: what changes and what stays

The skill removes comments that restate the code and keeps comments that explain a non-obvious rule. The first example below is a comment the skill would remove. The second is a pair of comments it would keep, taken from the skill's own example of content that must stay.

• Removed: a comment that says Initialize the variable above a declaration of a count variable set to zero
• Replaced: a banner like a line of equals signs around the word Authentication becomes one plain line, or is removed if the label adds nothing
• Kept: explanations of retry behaviour and duplicate handling for webhook events

Safety notes: scripts, network and file writes

This skill runs no scripts and uses no network access, according to its listing. Its allowed tools are Read, Write, Edit, Glob and Grep. It does not need to run commands to do its job.

The skill does write to files. It edits source files in place. The edits are limited to comments, but they are still writes, so do not use it where the agent must not touch files at all. Review the diff before committing, and check the skill's rules first if your codebase depends on specific comment formats that tools parse, because the skill may remove comments those tools expect.

• Runs scripts: no
• Uses network: no
• Edits files in place: yes, comments only
• Check first: comment formats read by tools or linters

Tips for getting good results

Run the skill on one file or module at a time so the diff is easy to review. Keep version control clean before the run so you can revert any change you do not want. Use the skill as a second pass after the agent writes code, and ask it to leave executable code alone in your request as well.

If a comment was legitimately valuable but got shortened, the skill's guidance says to keep the constraint itself in one line rather than the full history. Consider reviewing any comment that cites issue numbers or version history before accepting the change.

• Work on one file or module per run
• Commit or stash your work first so changes are easy to revert
• Review the diff before merging
• Check tool-parsed annotations before accepting removals