readme-designer
Skill2 starsChanged 16 days agoApache-2.0View on GitHub
Rewrite a README so a reader can tell what the project is and run it without scrolling, then check the result against a scoring rubric and fix every broken link. Cuts bloat, ornament and self-awarded badges; keeps the promise, the command, the output and the links. Use when creating, redesigning, modernizing, shortening, improving the readability of, or polishing a project README, repository documentation, or landing page.
In this repository19 other files it gives its agents.
- .agents/AGENTS.mdAGENTS.md
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- cross-platform-dev-scriptsSkill
- lessons-learntSkill
- multi-agent-docsSkill
- preflight-test-engineerSkill
- public-repo-release-reviewSkill
- readme-designerSkill
- saas-app-builderSkill
- third-party-skill-verifierSkill
- skill-creatorSkill
- cross-platform-dev-scriptsSkill
- lessons-learntSkill
- multi-agent-docsSkill
- preflight-test-engineerSkill
- public-repo-release-reviewSkill
- saas-app-builderSkill
- third-party-skill-verifierSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 52
Read first_command_line and line_count first.
- Distinctive: line 63
Delete decorative badges: Keep the ones a reader acts on: build status, licence, version.
- Distinctive: line 67
Delete the emoji that mark nothing: When every heading has one, none of them is a signal.
- Very specific: line 72
If the archived original is worth keeping, keep it -- docs/archive/ with a note at the top, links rewritten for the new location.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 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
README Designer
Make a README a reader finishes.
What this optimises for
A README is read by someone deciding, in about twenty seconds, whether this project is worth their afternoon. They want to know what it is, see it run, and find the rest. Everything that delays those three things is cost.
This skill previously optimised for the opposite. Its rubric awarded points
for <br/> tags, hero banners, badge rows, Mermaid diagrams and callouts,
and rubrics get optimised against: the result was a 736-line README that
scored A+ while the one command a reader needed sat at line 470. Version 2
pays for orientation, substance, restraint, brevity and working links, and
pays nothing for looks.
The rubric
scripts/audit_readme.py scores out of 100. Every number below is a
deliberate opinion, not a measurement:
| Dimension | Points | What earns them |
| Orientation | 20 | An # H1; a plain sentence saying what this is within 15 lines; the first runnable command within 40 lines. |
| Substance | 25 | A command block; real output; an install or usage heading; a licence; how to test and contribute. |
| Restraint | 20 | Few or no <br/> tags, at most five badges, emoji under 8 per 100 lines, no self-awarded grade badges. |
| Brevity | 15 | 200 lines or fewer. 350 is acceptable; past 600 it is a manual, not a README. |
| Links | 20 | Every relative path exists and every anchor matches a header. Broken links are the only hard error. |
--strict exits 1 unless the file scores 80 or more with no broken link.
Workflow
1. Audit before touching anything
python skills/readme-designer/scripts/audit_readme.py --target README.mdRead first_command_line and line_count first. They usually explain the
score on their own.
2. Cut before you write
Most bad READMEs are not badly written, they are too long. In order:
- Move reference material to
docs/and link it. Repository trees, full CLI catalogues, specification tables, architecture essays. - Delete every
<br/>. Markdown already separates blocks. The tags are layout by padding and cost the reader scrolling. - Delete decorative badges. Keep the ones a reader acts on: build status, licence, version. A badge the project awards itself -- a grade, a score, "production ready" -- reads as evidence and is not; the auditor warns about these by name.
- Delete the emoji that mark nothing. When every heading has one, none of them is a signal.
- Delete diagrams that restate the text. A diagram earns its place by showing something prose cannot: a real flow, a real shape.
If the archived original is worth keeping, keep it -- docs/archive/ with a
note at the top, links rewritten for the new location.
3. Structure what remains
# Project Name
One or two sentences: what this is, who it is for. No tagline in italics.
## The thing to try first
The single command, then its real output in a ```text block.
## Install / Usage
The other commands, with what they print.
## How it works
Two or three paragraphs. Longer goes to docs/ with a link.
## Contributing
How to run the tests and send a change.
## LicenseOrder by what a reader needs, not by what the project is proud of. If the project has one thing worth trying, put a real run of it -- real command, real output -- above everything else.
4. Write honestly
- Claims match evidence. "Checks passing" with a CI badge, not "Grade A" from your own tool. If a number is not measured, do not print it.
- Say what is unfinished. One line of "work in progress, here is what is and is not verified" buys more trust than any badge.
- Cut adjectives. "Ultra-modern, world-class, blazing-fast" is noise a reader discounts. The command and its output are the argument.
5. Re-audit until it passes
python skills/readme-designer/scripts/audit_readme.py --target README.md --strictFix broken links first -- they are the only defect here that is unambiguously wrong. Then read the top recommendations; each names the reader cost.
Starting from nothing
python skills/readme-designer/scripts/scaffold_readme.py \
--name "Project Name" --tagline "A tool that does one thing well" \
--output README.mdThe template is deliberately short and scores 100 on the rubric above. It is meant to be edited down, not up: a template that ships placeholders for a banner, a badge wall and a diagram gets them filled in rather than deleted.
What this skill will not do
- It will not add decoration to raise a score. If a change would not help a reader, it does not belong in the file.
- It will not invent metrics, grades or endorsements to fill a badge row.
- It will not delete content it cannot see a home for. Long material moves to
docs/and gets linked; it is not silently dropped.
Auditing other Markdown
The rubric assumes a README. Pointed at a guide or a specification, the brevity and orientation checks are the wrong shape -- a specification is supposed to be long. Use the link-integrity result, which is universal, and read the rest as advice rather than as a grade.
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.

