[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7EX204719QH2DT431C7PMS5":3,"skill-same:01M7EX204719QH2DT431C7PMS5":45},{"skillNo":4,"name":5,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":26,"repositoryHtmlUrl":27,"repositoryHomepageUrl":6,"ownerHtmlUrl":28,"ownerBlogUrl":29,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7EX204719QH2DT431C7PMS5","rules-checkup",null,"etherfunlab\u002Fworkflow","etherfunlab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F268832592?v=4","01M7EX1G4TQWFYMM15QWYN2S7F","EtherFun Lab","01M7EX1G5E2EB84BQTZ6AMKQWR",76,"MIT","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fmodules\u002Fcheckup\u002Fagent\u002Fskills\u002Frules-checkup\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: rules-checkup\ndescription: Use when the user asks to check up, audit or prune their rule files (CLAUDE.md, AGENTS.md, .claude\u002Frules, memory) — 体检一下, \"rules checkup\" — or after a reminder that Claude Code changed version. Takes a scope: `global` (only `~\u002F.claude`), `project` (default: everything this session loads) or `repo` (only what the repository commits). Produces a list of candidate deletions and edits; changes nothing until the user approves.\n---\n\n# Rules checkup\n\nRule files are written against the harness of the day. A new harness version can change the system prompt, so a rule that once filled a gap may now repeat the system prompt, fight it, or point at files that no longer exist. This skill lists what to cut. The user decides.\n\n## Three scopes\n\nThe scope comes as the skill's argument or in the user's words: `rules-checkup repo`, \"rules-checkup:repo\", 「仓内体检」. With none given, the scope is `project`.\n\n| Scope | Covers | Whose rules |\n|---|---|---|\n| `global` | Only the user-level rule files: `~\u002F.claude\u002FCLAUDE.md` and `~\u002F.claude\u002Frules\u002F` | The user's, in every project they open |\n| `project` | Every rule file this session loads, from `~\u002F.claude` down to the working directory, and memory | This user's, working here |\n| `repo` | Only what the repository commits, at any depth | Everyone's who clones it |\n\nRecommended order: `global`, then `project`, then `repo`. A checkup can move a rule down a layer, and the next pass then reads that layer in its final form.\n\nCandidates come only from the files in scope. State the scope in the first line of the output.\n\n## Read everything first\n\nIn every scope, also read your own current system prompt and tool descriptions, as you see them in this session. \"System prompt\" here means the harness's own text. Rule files and memory that the harness loaded into your context are not part of it, even though they sit beside it.\n\nIn `global` and `repo`, judge Redundant and Conflicting only against that system prompt and the files in scope. A rule that another layer happens to repeat stays: the other layer is not there for every project, or for every clone.\n\n### `global`\n\n1. `~\u002F.claude\u002FCLAUDE.md`, every `.md` file under `~\u002F.claude\u002Frules\u002F`, and every file they import with `@`.\n\nA global rule is paid for in every project. One more test applies here: a rule that names a single repository, project or product is a candidate to move down into that project. List these under a heading of their own, \"Belongs in one project\", after the six kinds.\n\n### `project`\n\n1. User level: `~\u002F.claude\u002FCLAUDE.md`, every `.md` file under `~\u002F.claude\u002Frules\u002F`, and every file they import with `@`.\n2. Project: every CLAUDE.md, `.claude\u002FCLAUDE.md` and CLAUDE.local.md in the working directory and in each directory above it, and their `@` imports. Do not stop at the repo root; Claude Code loads the ones above it too.\n3. AGENTS.md, only where your harness loads it. Claude Code loads one only when a CLAUDE.md imports it. A harness that reads AGENTS.md natively loads it from the repo root down to the working directory. An AGENTS.md this session does not load belongs to the `repo` scope.\n4. Every `.md` file under the project's `.claude\u002Frules\u002F`, subdirectories included.\n5. Memory: the index (`~\u002F.claude\u002Fprojects\u002F\u003Cproject>\u002Fmemory\u002FMEMORY.md`, or `.memory\u002FMEMORY.md`) and every memory file it points to.\n\n### `repo`\n\nWork from the repo root (`git rev-parse --show-toplevel`) and read only tracked files (`git ls-files`):\n\n1. Every CLAUDE.md and AGENTS.md at any depth, including directories this session has not loaded, and their `@` imports.\n2. Every `.md` file under `.claude\u002Frules\u002F`, subdirectories included.\n3. `.memory\u002F`, when the repo tracks one: the index and every memory file it points to.\n\nDo not read `~\u002F.claude`, the directories above the repo, or untracked files: a clone has none of them. Two tests change accordingly:\n\n- Duplicate counts only inside the repo. A rule that repeats one living outside the repo stays, because the repo's rules must hold on their own.\n- Stale fact also covers anything that exists only on this machine: an `@` import, path or script outside the repo that a rule depends on.\n\n## Six kinds of candidate\n\n| Kind | Test |\n|---|---|\n| Redundant | The system prompt already says the same thing, in the same direction |\n| Conflicting | The rule pushes against the system prompt or against another rule |\n| Stale fact | A path, command, file, flag or name in the rule no longer exists. Check each one with a command (`ls`, `command -v`, `git ls-files`, `--help`); never judge from memory |\n| Duplicate | The same rule lives in memory and in a rule file, in two rule files, or twice in one file. For a memory: the repo or its git history already records the same thing |\n| Hidden brake | A stop-or-ask rule fires only after the agent first classifies the situation (\"if this is a side task, ...\"). The moment it is needed is the moment the classification gets skipped |\n| Incident residue | Dates, PR or issue numbers, line numbers, measured figures, post-mortems or tombstones inside a rule |\n\n## Output\n\nOne numbered list grouped by kind. Each item: `file:line`, the rule (quoted, shortened), why it is a candidate, and the proposed change (delete \u002F shorten to \"...\" \u002F move to ...). End with \"Which items should I apply?\" and stop.\n\n## Rules\n\n- Change nothing until the user approves. Approval is per item unless the user says \"all\".\n- A checkup only removes, shortens or moves rules. It never adds one.\n- When a `project` checkup ends, record it in `~\u002F.claude\u002Fworkflow-checkup.json` with `jq`, creating the file if needed: set `last_checkup` to today (YYYY-MM-DD) and `last_version` to the first word of `claude --version`. It has ended once the user has answered, whether they approved every item or none. Leave `last_version` alone when there is no `claude` command.\n\n## Red flags\n\n- \"These are obviously stale, I'll just delete them.\" → List first, then wait.\n- \"While I'm here I'll add a rule for X.\" → Not in a checkup.\n- \"I remember this path exists.\" → Run the check.\n- \"The user is in a hurry, so I'll apply everything.\" → Hurry is not approval.\n","# Rules checkup\n\nRule files are written against the harness of the day. A new harness version can change the system prompt, so a rule that once filled a gap may now repeat the system prompt, fight it, or point at files that no longer exist. This skill lists what to cut. The user decides.\n\n## Three scopes\n\nThe scope comes as the skill's argument or in the user's words: `rules-checkup repo`, \"rules-checkup:repo\", 「仓内体检」. With none given, the scope is `project`.\n\n| Scope | Covers | Whose rules |\n|---|---|---|\n| `global` | Only the user-level rule files: `~\u002F.claude\u002FCLAUDE.md` and `~\u002F.claude\u002Frules\u002F` | The user's, in every project they open |\n| `project` | Every rule file this session loads, from `~\u002F.claude` down to the working directory, and memory | This user's, working here |\n| `repo` | Only what the repository commits, at any depth | Everyone's who clones it |\n\nRecommended order: `global`, then `project`, then `repo`. A checkup can move a rule down a layer, and the next pass then reads that layer in its final form.\n\nCandidates come only from the files in scope. State the scope in the first line of the output.\n\n## Read everything first\n\nIn every scope, also read your own current system prompt and tool descriptions, as you see them in this session. \"System prompt\" here means the harness's own text. Rule files and memory that the harness loaded into your context are not part of it, even though they sit beside it.\n\nIn `global` and `repo`, judge Redundant and Conflicting only against that system prompt and the files in scope. A rule that another layer happens to repeat stays: the other layer is not there for every project, or for every clone.\n\n### `global`\n\n1. `~\u002F.claude\u002FCLAUDE.md`, every `.md` file under `~\u002F.claude\u002Frules\u002F`, and every file they import with `@`.\n\nA global rule is paid for in every project. One more test applies here: a rule that names a single repository, project or product is a candidate to move down into that project. List these under a heading of their own, \"Belongs in one project\", after the six kinds.\n\n### `project`\n\n1. User level: `~\u002F.claude\u002FCLAUDE.md`, every `.md` file under `~\u002F.claude\u002Frules\u002F`, and every file they import with `@`.\n2. Project: every CLAUDE.md, `.claude\u002FCLAUDE.md` and CLAUDE.local.md in the working directory and in each directory above it, and their `@` imports. Do not stop at the repo root; Claude Code loads the ones above it too.\n3. AGENTS.md, only where your harness loads it. Claude Code loads one only when a CLAUDE.md imports it. A harness that reads AGENTS.md natively loads it from the repo root down to the working directory. An AGENTS.md this session does not load belongs to the `repo` scope.\n4. Every `.md` file under the project's `.claude\u002Frules\u002F`, subdirectories included.\n5. Memory: the index (`~\u002F.claude\u002Fprojects\u002F\u003Cproject>\u002Fmemory\u002FMEMORY.md`, or `.memory\u002FMEMORY.md`) and every memory file it points to.\n\n### `repo`\n\nWork from the repo root (`git rev-parse --show-toplevel`) and read only tracked files (`git ls-files`):\n\n1. Every CLAUDE.md and AGENTS.md at any depth, including directories this session has not loaded, and their `@` imports.\n2. Every `.md` file under `.claude\u002Frules\u002F`, subdirectories included.\n3. `.memory\u002F`, when the repo tracks one: the index and every memory file it points to.\n\nDo not read `~\u002F.claude`, the directories above the repo, or untracked files: a clone has none of them. Two tests change accordingly:\n\n- Duplicate counts only inside the repo. A rule that repeats one living outside the repo stays, because the repo's rules must hold on their own.\n- Stale fact also covers anything that exists only on this machine: an `@` import, path or script outside the repo that a rule depends on.\n\n## Six kinds of candidate\n\n| Kind | Test |\n|---|---|\n| Redundant | The system prompt already says the same thing, in the same direction |\n| Conflicting | The rule pushes against the system prompt or against another rule |\n| Stale fact | A path, command, file, flag or name in the rule no longer exists. Check each one with a command (`ls`, `command -v`, `git ls-files`, `--help`); never judge from memory |\n| Duplicate | The same rule lives in memory and in a rule file, in two rule files, or twice in one file. For a memory: the repo or its git history already records the same thing |\n| Hidden brake | A stop-or-ask rule fires only after the agent first classifies the situation (\"if this is a side task, ...\"). The moment it is needed is the moment the classification gets skipped |\n| Incident residue | Dates, PR or issue numbers, line numbers, measured figures, post-mortems or tombstones inside a rule |\n\n## Output\n\nOne numbered list grouped by kind. Each item: `file:line`, the rule (quoted, shortened), why it is a candidate, and the proposed change (delete \u002F shorten to \"...\" \u002F move to ...). End with \"Which items should I apply?\" and stop.\n\n## Rules\n\n- Change nothing until the user approves. Approval is per item unless the user says \"all\".\n- A checkup only removes, shortens or moves rules. It never adds one.\n- When a `project` checkup ends, record it in `~\u002F.claude\u002Fworkflow-checkup.json` with `jq`, creating the file if needed: set `last_checkup` to today (YYYY-MM-DD) and `last_version` to the first word of `claude --version`. It has ended once the user has answered, whether they approved every item or none. Leave `last_version` alone when there is no `claude` command.\n\n## Red flags\n\n- \"These are obviously stale, I'll just delete them.\" → List first, then wait.\n- \"While I'm here I'll add a rule for X.\" → Not in a checkup.\n- \"I remember this path exists.\" → Run the check.\n- \"The user is in a hurry, so I'll apply everything.\" → Hurry is not approval.\n",{},"main","eec36e50a538ad0e02620fb07c2be53851d51ce8","modules\u002Fcheckup\u002Fagent\u002Fskills\u002Frules-checkup",17,"https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow","https:\u002F\u002Fgithub.com\u002Fetherfunlab","https:\u002F\u002Fetherfun.xyz\u002F","2026-11-14 10:45:42",{"source":7,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add etherfunlab\u002Fworkflow --skill rules-checkup",{"tool":37,"command":38},"bunx","bunx skills add etherfunlab\u002Fworkflow --skill rules-checkup",{"tool":40,"command":41},"pnpm","pnpm dlx skills add etherfunlab\u002Fworkflow --skill rules-checkup","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fmodules\u002Fcheckup\u002Fagent\u002Fskills\u002Frules-checkup\u002FSKILL.md",true,[],[46,53,60,67,74,81],{"skillNo":47,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":18},"01M7EX3M6H9KPAXBR8TBA44JB3","doc-review","Use when reviewing docs in this repo (team version).","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Fcollisions\u002F.claude\u002Fskills\u002Fdoc-review\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":18},"01M7EX3G9ZYFKGD74DXKDWGT84","zh-hant-converter","Use when Simplified Chinese text needs a Traditional Chinese version — internal team docs converted with opencc s2hk, or a product's Simplified Chinese UI copy mirrored into its Traditional Chinese locale — or when the user says 转繁, 同步繁体, or \"convert to zh-Hant\".","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fskills\u002Fzh-hant-converter\u002Fagent\u002Fskills\u002Fzh-hant-converter\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":18},"01M7EX3B83SRTSBWQ3B6MKWKTF","cleaning-dev-caches","Use when this Mac's disk is nearly full, or the user asks to clear re-downloadable caches (「清缓存」「硬盘满了」, \"clear caches\") — Homebrew, npm \u002F pnpm \u002F yarn \u002F bun \u002F uv \u002F pip, Gradle, Playwright browsers, Edge and VS Code updater caches, simulator dyld caches, unused Rust toolchains. Not for build output (cleaning-dev-build-artifacts) or personal files.","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fskills\u002Fmacos-disk-cleanup\u002Fagent\u002Fskills\u002Fcleaning-dev-caches\u002FSKILL.md",[],[],{"skillNo":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":18},"01M7EX3647E2TEZ9TRDY689E3E","cleaning-dev-build-artifacts","Use when this Mac's disk is nearly full, or the user asks to clean build output — Rust target\u002F dirs or Claude Code worktrees (.claude\u002Fworktrees) under {{DEV_ROOT}} (「清编译产物」「清 target」「清 worktree」「硬盘满了」, \"clean build artifacts\"). Not for package-manager or app caches (cleaning-dev-caches) or personal files.","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fskills\u002Fmacos-disk-cleanup\u002Fagent\u002Fskills\u002Fcleaning-dev-build-artifacts\u002FSKILL.md",[],[],{"skillNo":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":18},"01M7EX2WMKSRYJT1AT02N5FNGV","ios-keyboard-page-shift","Use when, on iPhone Safari or an iOS WebView, a page or a full-screen app shell (top bar, scrolling list, bottom input) moves up when a text field is focused and stays up after the keyboard closes, leaving the top bar off-screen, the input mid-screen and an empty band below it, or when an earlier fix for that symptom did not hold.","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fskills\u002Fios-keyboard-page-shift\u002Fagent\u002Fskills\u002Fios-keyboard-page-shift\u002FSKILL.md",[],[],{"skillNo":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":18},"01M7EX2RQQGFSDGJS41KNYKWA4","codex-review","Use when the user asks to review a branch, PR or diff with codex (\"review this with codex\", 用 codex 审一下, \"codex review PR N\"). Runs `codex exec` from the shell with a hand-written, repo-aware prompt; falls back to one OpenRouter GPT call only when codex is out of quota. Covers the invocation contract and the failure modes that make codex look dead or silently produce no report.","https:\u002F\u002Fgithub.com\u002Fetherfunlab\u002Fworkflow\u002Fblob\u002Fmain\u002Fmodules\u002Freview\u002Fagent\u002Fskills\u002Fcodex-review\u002FSKILL.md",[],[]]