菜单
Reclaim disk space on a Mac dev machine by deleting regenerable dev caches (Cargo/zfb build output, node_modules, pnpm/npm caches, OS/app caches) and thinning APFS local snapshots so the space actually frees. Manual-only — invoke with /dev-clean-mac when a Mac is low on disk.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
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。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/Takazudo/claude-resources --skill dev-clean-mac命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
基于 SOC 职业分类
正在显示 SKILL.md
| name | dev-clean-mac |
| description | Reclaim disk space on a Mac dev machine by deleting regenerable dev caches (Cargo/zfb build output, node_modules, pnpm/npm caches, OS/app caches) and thinning APFS local snapshots so the space actually frees. Manual-only — invoke with /dev-clean-mac when a Mac is low on disk. |
| disable-model-invocation | true |
| argument-hint | [--keep-node-modules] |
Frees disk on a Mac dev machine by deleting caches and build artifacts that
regenerate on demand. Tuned for this layout — $HOME/repos (with myoss/zfb),
a shared Cargo target dir, pnpm/npm caches — but every step is existence-guarded,
so it runs safely on any Mac with a similar structure (e.g. a second machine).
Why a script and not ad-hoc rm: on APFS, deleting files frees nothing while
local Time Machine snapshots still reference the data — df keeps showing the old
number. The script ends by thinning those snapshots (tmutil thinlocalsnapshots),
which is what makes the reclaimed space materialize. Skipping that step is the #1
reason a cleanup "does nothing."
The script lives at scripts/clean-mac.sh in this skill directory.
Dry run first — show the user what will be freed, delete nothing:
bash "$HOME/.claude/skills/dev-clean-mac/scripts/clean-mac.sh" --dry-run
Relay the per-path sizes and the measured total to the user.
Confirm node_modules. By default the script deletes all node_modules
under $HOME/repos (reinstallable via pnpm/npm install). If the user is
mid-work on a project, pass --keep-node-modules to preserve them.
Run for real:
bash "$HOME/.claude/skills/dev-clean-mac/scripts/clean-mac.sh" --run
# or, to keep node_modules:
bash "$HOME/.claude/skills/dev-clean-mac/scripts/clean-mac.sh" --run --keep-node-modules
The deletions can take a few minutes (millions of small files); it may
auto-background. Report the BEFORE/AFTER df lines it prints.
Cargo: the shared target-dir from $HOME/.cargo/config.toml (defaults to
$HOME/.cargo-target, dominated by zfb), $HOME/.cargo/registry, and every
per-project target/ under $HOME/repos (detected by cargo's CACHEDIR.TAG).
zfb: checked-in release tarballs (zfb-*.tar.gz) and all .zfb-build/ dirs.
node_modules: every one under $HOME/repos (unless --keep-node-modules).
JS caches: pnpm store prune, npm cache clean, $HOME/Library/Caches/pnpm.
OS/app caches: ms-playwright, electron, copilot, $HOME/.cache/codex-runtimes,
Homebrew cache, Adobe/CreativeCloud logs, unavailable iOS simulators.
Downloads: *.dmg / *.pkg installers (re-downloadable).
APFS local snapshots: thinned at the end to release everything above.
Source trees, image/data stashes (e.g. zmod-image-stash), browser profiles,
and app data (Claude/Chrome/Slack containers) — only regenerable caches and
artifacts. Every path is verified to be under $HOME before removal; missing
paths are skipped silently, which is what makes it portable across machines.
First builds/installs after running are slower (caches re-download, projects
rebuild) — that's the expected trade for the space.
Safe to re-run anytime; all steps are idempotent.
Local Time Machine snapshots are convenience restore points, not your real
backup — thinning them does not affect Time Machine backups on disk/network.
Get a second opinion from OpenAI Codex CLI on a plan or approach. Use when: (1) During planning phase of /x-as-pr or /x-wt-teams to validate the approach, (2) User says 'codex 2nd', 'second opinion', or 'codex opinion', (3) Wanting an alternative perspective before committing to a plan. Sends context and plan to codex, returns feedback. Called automatically by /x-as-pr and /x-wt-teams during planning.
Generate or redesign raster images (PNG mockups, illustrations, photos, UI design polish) from inside Claude Code by driving Codex CLI's built-in $imagegen, billed to ChatGPT-included usage instead of per-image OpenAI API charges. Use whenever the user wants to create/generate an image, make a mockup or illustration, produce a photo, polish or redesign a UI from a screenshot, or get a PNG out of a session — even if they don't mention Codex or "imagegen". This is the preferred image-generation path because Anthropic models cannot generate raster images and this route avoids API billing.
Web research using OpenAI Codex CLI (codex exec). PREFERRED over general web research. Use when: (1) User says 'research', 'codex research', 'look up', or 'investigate', (2) Researching libraries, APIs, best practices, or technical topics, (3) Gathering information from the web. Codex performs research, Claude Code synthesizes. Falls back to Claude Code researcher subagent if codex unresponsive.