repo-polisher
Skill0 starsChanged 5 months agoMITView on GitHub
Audit and modernize a code repository end to end: fix real bugs, upgrade legacy language/runtime usage, add missing project scaffolding (dependency manifest, .gitignore, tests, CI), and write a polished README. Use this whenever the user wants to "clean up", "modernize", "polish", "improve", "refactor", or "open-source" a repo or project, asks to make a project "look professional" or "presentable" for recruiters/portfolios, points at an old or messy codebase, or shares a GitHub repo and wants it improved — even if they don't say the word "polish".
In this repository10 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- code-reviewerSkill
- commit-and-prSkill
- competitive-analysisSkill
- debuggerSkill
- prd-writerSkill
- security-reviewSkill
- tech-design-docSkill
- test-authorSkill
- user-storiesSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 7 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
Read the actual code, not just the file names.
- Distinctive: line 32
Do not start editing until you can state, in one sentence, what the project does and how someone runs it.
- Distinctive: line 80
Run the code and the tests one more time.
- Distinctive: line 87
Prefer the smallest change that achieves the goal.
- Distinctive: line 88
Never invent results, benchmarks, or features in the README.
- Distinctive: line 89
If you can't run the project in the environment, say so and reason carefully instead of guessing.
- Distinctive: line 90
Keep the original authors' intent and authorship intact unless asked otherwise.
≈ “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
Repo Polisher
Take a rough, dated, or undocumented repository and turn it into something a maintainer would be proud to publish. The goal is a repo that a stranger can understand in two minutes and run in five.
Work in this order. Each phase builds on the last, and doing them out of order (e.g. rewriting the README before you understand the code) produces shallow, inaccurate results.
1. Understand before you touch anything
Read the actual code, not just the file names. Identify the language(s), the entry points, how the project is run, what it depends on, and what it is actually trying to do. Note the state of the basics: is there a dependency manifest, a README, tests, a license, a .gitignore? Skim the commit history if it is available.
Do not start editing until you can state, in one sentence, what the project does and how someone runs it. If you cannot, keep reading.
2. Find the real bugs
Surface-level tidying is not the value here — correctness is. Look for genuine problems, not just style: code that wouldn't run on a current runtime, hardcoded absolute paths, broken data loading, silent failures, off-by-one logic, resource leaks, and incorrect assumptions about inputs. When you find a bug, verify it (run the code or the relevant path) before claiming it, and verify your fix after.
Be honest about severity. A misleading data pipeline that makes results wrong matters far more than inconsistent quote style.
3. Modernize
Bring the code up to current, idiomatic practice for its language without gold-plating. Typical moves: migrate off end-of-life language versions, replace deprecated APIs, remove dead code, extract repeated logic, add type hints or signatures where they aid clarity, and make configuration explicit instead of hardcoded. Keep the diff focused — the point is a codebase the author can still recognize, not a rewrite.
4. Add the missing scaffolding
A repo that can't be installed or trusted won't be used. Add whatever is missing from the project's hygiene baseline:
- Dependency manifest (
requirements.txt,package.json,go.mod,Cargo.toml, etc.) .gitignoreappropriate to the stack- A test or two that actually exercise the core path (even one meaningful test signals quality)
- A license, if the author wants one
- Basic CI (e.g. a GitHub Actions workflow that installs and runs tests) when it fits
See references/checklist.md for a per-language scaffolding checklist.
5. Write the README last
Now that the code is correct and runnable, write a README that reflects reality. A strong README opens with a one-line description of what the project does, then covers: why it exists, how to install it, how to run it (copy-paste commands that actually work), and a short note on results or output where relevant. Keep it honest — if the dataset is tiny or the model is a baseline, say so. Overclaiming reads as junior.
6. Verify and summarize
Run the code and the tests one more time. Then tell the user what you changed and why, grouped by impact (bugs fixed, then modernization, then scaffolding, then docs) — they have to defend these changes, so give them the reasoning, not just a file list.
Guardrails
- Prefer the smallest change that achieves the goal. Resist rewriting working code for taste.
- Never invent results, benchmarks, or features in the README. Verify claims against the code.
- If you can't run the project in the environment, say so and reason carefully instead of guessing.
- Keep the original authors' intent and authorship intact unless asked otherwise.
Example
Input: "Here's my old college project, can you make it look good for my GitHub?"
Output: A repo where the Python 2 script now runs on Python 3, a data-loading
bug that silently trained on the wrong column is fixed, a requirements.txt,
.gitignore, and one real test are added, and the README explains what it does,
how to run it, and honest accuracy numbers — plus a short changelog of what was
fixed and why.
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.

