菜单
Audit and tidy a repo or working folder: fix layout drift, consolidate misplaced files, remove stale artifacts, align docs with disk. Use when organizing any project, scratch folder, or new repo — before release or after a refactor.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | repo-hygiene |
| description | Audit and tidy a repo or working folder: fix layout drift, consolidate misplaced files, remove stale artifacts, align docs with disk. Use when organizing any project, scratch folder, or new repo — before release or after a refactor. |
Keep a tree navigable, documented, and free of stale cruft — without breaking active work or deleting anything the user still needs.
Complements code-writing (implementation) and code-reviewer (diff quality).
Do not prescribe a fixed layout. Derive structure from the project type and
whatever the repo already documents (README, CLAUDE.md, package manifest,
framework defaults).
find . -maxdepth 4 -not -path './.git/*' | sort
git status 2>/dev/null || true
Read README, CLAUDE.md, pyproject.toml / package.json / Cargo.toml if
present. Compare disk vs documented intent. List issues:
| Issue type | Example |
|---|---|
| Path drift | README says src/ but code lives at repo root |
| Duplicate trees | Old copy left after a move |
| Empty scaffolds | Placeholder dirs with no files |
| Stale outputs | Old run artifacts, caches, build dirs |
| Mixed concerns | Tests, scripts, and results all in one flat folder |
| Local-only files | Editor config, large data that should be gitignored |
Start from what the repo is, not a template from another project:
| Context | Typical anchors (adapt, don't copy blindly) |
|---|---|
| Library / package | src/ or <pkg>/, tests/, docs/, root README |
| App / service | app code, config, tests/, deploy/infra if any |
| Research / scratch | scripts/, notebooks/ or doc/, results/ or outputs/ (often gitignored) |
| Monorepo | one top-level dir per package; shared docs at root |
| Pre-git experiment | minimal: README stub, .gitignore, one clear output dir |
Principles (always):
src/ layout, Node package.json at root, etc..gitignore for junk,
delete failed experiments rather than accumulating old/, backup/, test2/.If the repo has no stated convention yet, propose the smallest layout that fits the work — do not import structure from unrelated repos (plugins, monorepos, etc.).
| Safe to delete (usually) | Confirm with user first |
|---|---|
| Empty directories | Anything under results/, outputs/, benchmarks/, runs/ |
__pycache__/, .pytest_cache/, *.egg-info/, node_modules/ (reinstallable) | Plots, checkpoints, or logs that might be the only record of a run |
| Duplicate copies superseded by a move | WIP code still referenced in docs or scripts |
| Editor-local config not meant to be shared | Data files the user still needs |
Never delete without checking: uncommitted work, the only copy of a file, or paths referenced in README, CI, imports, or scripts.
git mv when tracked)..gitignore.git status is clean or changes are intentional.When reporting a hygiene pass:
## Repo hygiene report
### Context
- Project type: ...
- Conventions read from: ...
### Layout (before → after)
- ...
### Removed
- ...
### Moved
- ...
### Docs updated
- ...
### Left alone (and why)
- ...
### Follow-ups
- ...
| Approach | Use when |
|---|---|
This skill (repo-hygiene) | Inline cleanup; need a checklist |
| repo-organizer subagent | Large tree or many deletes; audit first, apply after confirmation |
Do not fold this into code-writing or code-reviewer — those focus on code diffs, not tree structure.
Deep-read an academic paper and produce a durable, critical research note — not a loose summary. Use when given a paper title, DOI, arXiv ID, publisher/PDF/GitHub URL, or local PDF, or when asked to read closely, summarize, extract contributions, analyze experiments, critique, compare to related work, reproduce results, or maintain a paper-reading repository index. Works across fields (astro/physics, ML, bio, security, …); examples lean astro-ph but the workflow is general.
Remote, slice-level access to FLAMINGO HDF5 files on COSMA via the `hdfstream` Python module — no need to download whole snapshots. Covers install, server connection, navigation, dataset slicing, multi-slice batched reads, and integration with swiftsimio. Use when reading any FLAMINGO data product remotely.
FLAMINGO past-lightcone products: virtual observers, particle lightcones, HEALPix shell maps, integrated_maps, and halo lightcones. Use when building mock surveys, stacking Compton-y or lensing maps, or cross-correlating with observations.
Review code changes against requirements and quality standards before merge. Use when dispatching a review subagent, after major features, before merging, or when the user asks for a code review. Requires reading the actual diff.
Style and structure for research manuscript text: IMRaD, hourglass narrative, formal tone, topic sentences, reproducible Methods, captions that tell the story. Use when drafting or editing titles, abstracts, sections, captions, or reports.
Iterative VLM review of compiled LaTeX PDFs for layout clashes (text vs figures, tables, long equations). Use after the first compile or any float/equation edit; loop fix-recompile-recheck until no clashes remain.