repolish
Skill4 starsChanged 13 days agoMITView on GitHub
Review and clean an entire repository for readability and maintainability, standardize language formatting, refactor while preserving behavior, and remove proven-unused code using coordinated subagents and resumable file accounting. Use for repository-wide cleanup or organization requests; a small feature or isolated formatting fix does not require a whole-repository pass.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- asale-ai/anything-to-skilltheir own version, first seen Aug 29, 2026
- asale-ai/seo-geo-skilltheir own version, first seen Aug 29, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- asale-ai/seo-geo-skill★ 6same name
- asale-ai/anything-to-skill★ 5same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈6
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 10
Read references/workflow.md before starting.
- Distinctive: line 16
Remove files or code only when evidence supports non-use.
- Distinctive: line 24
Map and baseline: Read applicable instructions, inspect working changes, discover workspaces, languages, entrypoints, generators, public …
- Distinctive: line 32
Use descriptive names, focused functions, straightforward control flow, and blank lines between logical steps.
- Distinctive: line 36
Do not split semicolons, braces, strings, or comments using text replacement to simulate formatting.
- Distinctive: line 42
Use available subagent capacity only where work can be separated.
- Very specific: line 46
Use scripts/inventory.py to initialize, reconcile, record dispositions, and report coverage.
≈ “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
Repolish
Turn a repository into code people can comfortably read and maintain. Account for every file in the agreed scope; rewrite a file only when the change improves it. Preserve features, observable behavior, public contracts, and user work.
Read references/workflow.md before starting. It defines discovery, formatter selection, worker assignments, deletion evidence, verification, and the inventory helper commands.
Defaults
- Perform full behavior-preserving cleanup: formatting, naming, control flow, function responsibilities, module organization, duplication, and dependency boundaries.
- Standardize each language with a conventional formatter, even where existing styles conflict. Add minimal pinned development tooling if needed; preserve runtime dependencies, compiler semantics, and language targets.
- Remove files or code only when evidence supports non-use. Leave uncertain candidates in place and explain them.
- Continue past pre-existing failures, record them, and fix cleanup regressions. Missing checks and unverified behavior remain explicit gaps.
- Work locally. Cleanup does not authorize automatic commits, publication, deployment, runtime dependency upgrades, or public API changes.
Honor narrower user scope and repository instructions. Do not turn a small request into a repository overhaul. When cleanup is requested, carry out justified local changes without repeated confirmation; a request only to plan or review does not authorize implementation.
Execute in passes
- Map and baseline. Read applicable instructions, inspect working changes, discover workspaces, languages, entrypoints, generators, public consumers, and existing checks. Create the external file ledger and capture baseline results.
- Format centrally. Choose one primary formatter per language or compatible file domain. Normalize indentation, wrapping, and logical spacing with syntax-aware tools. Keep mechanical changes distinguishable from structural edits; check the formatted baseline before parallel work.
- Review in parallel. Partition by cohesive subsystem with exclusive file ownership. Each worker reads its files and relevant callers, makes justified improvements, and runs focused checks. Formatting alone is not code review.
- Integrate. The lead resolves shared-file requests, cross-boundary moves, imports, configuration, and deletions. Run relevant combined checks and inspect the final diff for behavior changes and unrelated edits.
- Reconcile and finish. Refresh inventory and receipts against final contents. Account for additions and removals. Finish only when there are no pending files or unexplained regressions; label blocked coverage or verification limits as incomplete.
Readability and fundamentals
Use descriptive names, focused functions, straightforward control flow, and blank lines between logical steps. Prefer useful intent comments over narration of obvious operations. Preserve nuanced validation, error handling, ordering, state transitions, resource cleanup, and async behavior.
Apply SOLID where responsibilities, substitutability, interfaces, or dependency direction cause concrete maintenance problems. Use DRY for shared knowledge, KISS for simpler solutions, and YAGNI to avoid speculative extension points. Cohesive functions or modules often suffice; do not manufacture classes, interfaces, wrappers, or a new architecture to satisfy a checklist.
Do not split semicolons, braces, strings, or comments using text replacement to simulate formatting. Do not force tabs where the language requires spaces, or expand intentionally compact data and generated output merely to look handwritten.
Parallel ownership
The lead owns the ledger, shared configuration, package manifests, lockfiles, common exports, cross-subsystem changes, integration, and completion claims. Workers own disjoint file sets, never spawn additional workers, and return proposals for edits outside their ownership. Serialize shared writers and checks that mutate the same outputs.
Use available subagent capacity only where work can be separated; otherwise run serially and disclose that limitation. A worker returning successfully is not evidence that the integrated repository works.
File accounting
Use scripts/inventory.py to initialize, reconcile, record dispositions, and report coverage. It writes only an external ledger and cannot rewrite or delete source. Only the lead writes the ledger; workers return receipts.
Record reviewed, changed, removed, excluded, or blocked with specific evidence. reviewed means the relevant content was actually inspected; changed includes final-content review, not just a formatter run. Generated, vendored, binary, ignored, linked, and nested-repository material needs an explicit exclusion reason. Read and review human-authored configuration, documentation, and generator sources where they affect the codebase.
The helper checks inventory freshness, not semantic correctness or test success. Refresh receipts after content changes. Preserve uncertain removal candidates and never report exclusions as refactored code.
Deliver
Report substantive improvements, files removed and why, coverage counts, checks with actual outcomes, baseline failures, retained candidates, and remaining gaps. Link the external ledger/report location. Distinguish final integration verification from worker checks and formatting. Describe unverified areas without claiming behavioral equivalence that the evidence cannot establish.
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.

