[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7586NAC5TPWNAABE2V2SYGB":3,"skill-same:01M7586NAC5TPWNAABE2V2SYGB":46},{"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":28,"ownerHtmlUrl":29,"ownerBlogUrl":28,"ownerDescription":30,"updatedAt":31,"install":32,"resources":45},"01M7586NAC5TPWNAABE2V2SYGB","corpus-sweep",null,"nubjs\u002Fnub","nubjs","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F6785695?v=4","01M7583N3EYYHSNB6FJQXCDMQH","Nub","01M7583N44MVMSN8385QEF3C8V",4440,"MIT","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fcorpus-sweep\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: corpus-sweep\ndescription: Run a large sharded measurement sweep over npm packages (the build-jail catalog probe, or any harness that installs thousands of package-versions and records a verdict per run). Invoke BEFORE launching a sweep, before believing any standalone reproduction of a sweep failure, and before reporting that a failure is or is not a real defect. Carries the self-deception patterns that each cost hours: an install exiting 0 while the build was silently skipped, `find` not following store symlinks, a restricted PATH swapping the toolchain out from under a probe, sequential arms sharing a warm store, and a batch query whose zero hits were never validated against a known positive. Also the shard mechanics, the disk and OS-indexing overheads, artifact cleanup, and the rule that a malicious package must never be executed.\n---\n\n# Running a corpus sweep\n\nA sweep installs thousands of third-party package-versions and records a verdict for each. The\nverdicts feed something real — for the build jail, the capability catalog that decides what actual\nuser installs are permitted. So a wrong verdict is not a flaky test, it is a shipped defect.\n\n**The single most important thing in this skill: nearly every wrong conclusion comes from a\nSTANDALONE REPRODUCTION THAT SILENTLY TESTED NOTHING.** The harness says a package fails. You run it\nby hand, it passes, and you conclude the harness is wrong. It is almost always the reverse.\n\n## ⛔ Before you believe any standalone reproduction\n\nRun this checklist. Each line is a measured failure that produced a confident wrong answer.\n\n| Check | Why |\n| --- | --- |\n| **Did the script actually RUN?** | `nub install` exits **0** while the trust policy skips the build entirely — `WARN ignored build scripts for N package(s)`. You must run `nub approve-builds --all` after the install. An exit code proves nothing about whether a lifecycle script executed. |\n| **Did you look for the artifact with `find -L`?** | Under the isolated linker every `node_modules` entry is a SYMLINK into the global store. Plain `find` does not follow symlinks and reports **zero** addons where `ls` shows them present. |\n| **What is on PATH?** | Restricting `PATH` to `\u002Fusr\u002Fbin` silently hands node-gyp Xcode's Python 3.9 instead of the host's 3.14 — so the Python-dependent failure you are chasing cannot reproduce. Print the resolved `python3`, `node`, and `node-gyp` versions in the probe output, not just the exit code. |\n| **Is each arm getting a FRESH home AND a fresh store?** | Sequential arms against a shared store make the second one warm. That confounds every A\u002FB, and it is how four consecutive wrong answers about one package were reached. |\n| **Are you varying exactly ONE thing?** | A separate tool (`classify-broken.sh`) differing from the harness in fixture, env scrubbing, pinned Node\u002FPython AND jail state is not a control for concurrency. |\n\n**A clean result where you expected a messy one is a signal to distrust the instrument, not to\nconclude.** Three passes in a row after a real failure means you probably are not running the thing\nyou think you are running.\n\n## ⛔ Validate any query or filter against a KNOWN POSITIVE\n\nA screen that returns zero hits has two explanations and you cannot tell them apart without a\ncontrol. Measured: an OSV screen over 2,250 entries returned 0 MAL-* hits. That happened to be\ncorrect — but the only way to know was to re-run it against a package known to be flagged\n(`@ctrl\u002Ftinycolor@4.1.2` → `MAL-2025-47141`) and confirm the instrument fires. Do this for every\nbatch API call, grep filter, and classifier before reading a conclusion off it.\n\nBatch APIs deserve a second control: put the known positive in the MIDDLE of a full-size batch and\ncheck it is still found at the right index. Silent truncation is real.\n\n## ⛔ Malicious packages: never execute, never catalog\n\nA package with a `MAL-*` advisory must never have its scripts run, on this machine or any machine.\n\n- The PM under test may refuse it correctly — but the **reference arms (npm, pnpm) have no OSV\n  screen** and typically run with `--dangerously-allow-all-scripts`. If a refusal verdict is\n  computed *after* the oracle arms, the harness executes the malicious script itself.\n- So: detect the refusal and **return before any oracle arm runs.** Verify the ordering in code,\n  not by assumption — the guarantee must not be incidental to line order.\n- Better still, screen the WORKLIST against OSV before the sweep starts and drop hits entirely, so\n  the tarball is never even fetched.\n- A refusal is its OWN verdict. Scoring it as a defect of the PM under test blames the tool for\n  working correctly, and lands the package in the wrong bucket of the final report.\n\n## Sharding\n\nRound-robin the worklist so each shard gets a mix of heavy head and cheap tail:\n\n```sh\npython3 -c \"\nlines=[l.strip() for l in open('worklist.txt') if l.strip()]\nfor i in range(3):\n    open(f'shard{i}.txt','w').write('\\n'.join(lines[i::3])+'\\n')\"\n```\n\nThen launch each shard as its OWN harness-tracked background task — never `&`, which the harness\nrejects because a detached job cannot report completion.\n\n- **Check fixture isolation first.** Concurrent shards are only safe if each process roots its\n  fixtures uniquely (`fs.mkdtempSync`). Verify before launching, not after.\n- 3 shards took throughput from ~0.7 to ~2 runs\u002Fmin on a 10-core box. More is not obviously better:\n  contention makes packages whose scripts run their OWN installer fail transiently.\n- **Records must be per-package files** so shards never write the same path.\n\n## The host will fight you\n\n- **Spotlight and Time Machine index the churn.** Each cell installs a full dependency tree into a\n  fresh `$HOME`, so a sweep generates millions of file events. Measured: `fseventsd` 60%,\n  `backupd` 30%, `mds` + `spotlightknowledged` 47% — about 1.7 cores of pure overhead. Fix with\n  `touch \u003Ccache>\u002F.metadata_never_index` and `tmutil addexclusion \u003Ccache>`, both idempotent.\n- **High load with slow progress is not always your processes.** Check `ps -Ao pid,%cpu,comm -r`\n  before assuming contention; the answer may be an OS daemon.\n- **Fixture roots leak on SIGKILL.** A harness that cleans up on normal exit does not clean up when\n  you `pkill` it — and stop-fix-restart is the normal loop when triaging. Measured: 31 orphans,\n  18.9 GB, on a disk already at 98%. Sweep anything untouched for >30 min at startup; a live shard\n  touches its root continuously so the threshold cannot catch one in use.\n- **Wall-clock numbers taken during a sweep are untrustworthy.** Use load-independent evidence.\n\n## Clean up when you are done\n\nA sweep and its debugging leave several distinct piles. Delete all of them:\n\n- the harness's own fixture roots (`mkdtemp` dirs under the cache);\n- any ad-hoc reproduction trees you created while triaging — these are the biggest surprise,\n  measured at **12 GB** from one night of hand-testing;\n- temp catalogs and worklists in `\u002Ftmp`;\n- per-cell logs for records you have finished analyzing, if the verdict is recorded.\n\nKeep the `results.json` records — they are small (~39 KB each, ~0.1 GB for a 2,250 corpus) and they\nare the actual output.\n\n## Stopping to fix the harness\n\nEvery harness fix invalidates the records taken before it, because a record means something\ndifferent under a changed instrument. That is what provenance hashes exist to expose.\n\n- **Purge exactly the records the fix could change**, not all of them, when you can characterise the\n  set (e.g. \"only runs pinned to Node ≤15\"). Purge everything when the change is corpus-wide.\n- **Never edit the harness while a batch runs.** The driver script is re-read on every package\n  spawn; editing it mid-run silently corrupted 54 of 100 packages once. Stop, fix, restart.\n- A binary rebuild mid-run is safe IF the runner snapshots the binary at startup. Confirm it does.\n- Weigh the trade honestly: a correct instrument with partial coverage beats a complete corpus\n  measured by an instrument you already know is wrong. But say the coverage number plainly.\n\n## Bringing the harness up on a NEW PLATFORM\n\n**Load the `probe-platforms` skill.** It owns the bring-up ladder (debug with the SMALLEST payload —\nsix Windows faults were found using a 25-minute package when seconds would have done), the Windows\nspawn\u002Fpath\u002Fdisk faults, the Linux Landlock and node-layout traps, and the remote-shell mechanics.\n\nTwo rules from it that bite during a sweep specifically:\n\n- **Shuffle the worklist before sharding.** It is name-sorted, so a contiguous slice hands every\n  shard the same heavy family at once — four shards once sat 12 minutes on one family and produced\n  zero records. A seeded shuffle fixed it in seconds.\n- **Confirm the override actually ENGAGED before believing any cell.** The catalog parser rejects a\n  malformed catalog (e.g. `read` alongside `write: \"disk\"`, which is redundant) and falls back to the\n  compiled-in one *silently* — so a run you believe grants network may have none.\n\n## Reading results\n\n- **Coverage first, intersected with the worklist.** Counting every record on disk inflates it with\n  earlier runs. Packages with NO record mean the recorded set is a biased sample — the heavy native\n  builds are exactly the ones that fail, so survivors are not the corpus.\n- **Read the FIRST error, never the tail.** These logs end in a stack trace and a summary; the cause\n  is ~40 lines earlier.\n- **Group by CAUSE before concluding.** A cluster of similar-looking strings is not evidence of a\n  shared cause; clustering by an error substring once merged two unrelated bugs.\n- **Frequency tables must count DISTINCT PACKAGES, not records.** One package measured at four\n  versions looks like ecosystem-wide leakage otherwise.\n","# Running a corpus sweep\n\nA sweep installs thousands of third-party package-versions and records a verdict for each. The\nverdicts feed something real — for the build jail, the capability catalog that decides what actual\nuser installs are permitted. So a wrong verdict is not a flaky test, it is a shipped defect.\n\n**The single most important thing in this skill: nearly every wrong conclusion comes from a\nSTANDALONE REPRODUCTION THAT SILENTLY TESTED NOTHING.** The harness says a package fails. You run it\nby hand, it passes, and you conclude the harness is wrong. It is almost always the reverse.\n\n## ⛔ Before you believe any standalone reproduction\n\nRun this checklist. Each line is a measured failure that produced a confident wrong answer.\n\n| Check | Why |\n| --- | --- |\n| **Did the script actually RUN?** | `nub install` exits **0** while the trust policy skips the build entirely — `WARN ignored build scripts for N package(s)`. You must run `nub approve-builds --all` after the install. An exit code proves nothing about whether a lifecycle script executed. |\n| **Did you look for the artifact with `find -L`?** | Under the isolated linker every `node_modules` entry is a SYMLINK into the global store. Plain `find` does not follow symlinks and reports **zero** addons where `ls` shows them present. |\n| **What is on PATH?** | Restricting `PATH` to `\u002Fusr\u002Fbin` silently hands node-gyp Xcode's Python 3.9 instead of the host's 3.14 — so the Python-dependent failure you are chasing cannot reproduce. Print the resolved `python3`, `node`, and `node-gyp` versions in the probe output, not just the exit code. |\n| **Is each arm getting a FRESH home AND a fresh store?** | Sequential arms against a shared store make the second one warm. That confounds every A\u002FB, and it is how four consecutive wrong answers about one package were reached. |\n| **Are you varying exactly ONE thing?** | A separate tool (`classify-broken.sh`) differing from the harness in fixture, env scrubbing, pinned Node\u002FPython AND jail state is not a control for concurrency. |\n\n**A clean result where you expected a messy one is a signal to distrust the instrument, not to\nconclude.** Three passes in a row after a real failure means you probably are not running the thing\nyou think you are running.\n\n## ⛔ Validate any query or filter against a KNOWN POSITIVE\n\nA screen that returns zero hits has two explanations and you cannot tell them apart without a\ncontrol. Measured: an OSV screen over 2,250 entries returned 0 MAL-* hits. That happened to be\ncorrect — but the only way to know was to re-run it against a package known to be flagged\n(`@ctrl\u002Ftinycolor@4.1.2` → `MAL-2025-47141`) and confirm the instrument fires. Do this for every\nbatch API call, grep filter, and classifier before reading a conclusion off it.\n\nBatch APIs deserve a second control: put the known positive in the MIDDLE of a full-size batch and\ncheck it is still found at the right index. Silent truncation is real.\n\n## ⛔ Malicious packages: never execute, never catalog\n\nA package with a `MAL-*` advisory must never have its scripts run, on this machine or any machine.\n\n- The PM under test may refuse it correctly — but the **reference arms (npm, pnpm) have no OSV\n  screen** and typically run with `--dangerously-allow-all-scripts`. If a refusal verdict is\n  computed *after* the oracle arms, the harness executes the malicious script itself.\n- So: detect the refusal and **return before any oracle arm runs.** Verify the ordering in code,\n  not by assumption — the guarantee must not be incidental to line order.\n- Better still, screen the WORKLIST against OSV before the sweep starts and drop hits entirely, so\n  the tarball is never even fetched.\n- A refusal is its OWN verdict. Scoring it as a defect of the PM under test blames the tool for\n  working correctly, and lands the package in the wrong bucket of the final report.\n\n## Sharding\n\nRound-robin the worklist so each shard gets a mix of heavy head and cheap tail:\n\n```sh\npython3 -c \"\nlines=[l.strip() for l in open('worklist.txt') if l.strip()]\nfor i in range(3):\n    open(f'shard{i}.txt','w').write('\\n'.join(lines[i::3])+'\\n')\"\n```\n\nThen launch each shard as its OWN harness-tracked background task — never `&`, which the harness\nrejects because a detached job cannot report completion.\n\n- **Check fixture isolation first.** Concurrent shards are only safe if each process roots its\n  fixtures uniquely (`fs.mkdtempSync`). Verify before launching, not after.\n- 3 shards took throughput from ~0.7 to ~2 runs\u002Fmin on a 10-core box. More is not obviously better:\n  contention makes packages whose scripts run their OWN installer fail transiently.\n- **Records must be per-package files** so shards never write the same path.\n\n## The host will fight you\n\n- **Spotlight and Time Machine index the churn.** Each cell installs a full dependency tree into a\n  fresh `$HOME`, so a sweep generates millions of file events. Measured: `fseventsd` 60%,\n  `backupd` 30%, `mds` + `spotlightknowledged` 47% — about 1.7 cores of pure overhead. Fix with\n  `touch \u003Ccache>\u002F.metadata_never_index` and `tmutil addexclusion \u003Ccache>`, both idempotent.\n- **High load with slow progress is not always your processes.** Check `ps -Ao pid,%cpu,comm -r`\n  before assuming contention; the answer may be an OS daemon.\n- **Fixture roots leak on SIGKILL.** A harness that cleans up on normal exit does not clean up when\n  you `pkill` it — and stop-fix-restart is the normal loop when triaging. Measured: 31 orphans,\n  18.9 GB, on a disk already at 98%. Sweep anything untouched for >30 min at startup; a live shard\n  touches its root continuously so the threshold cannot catch one in use.\n- **Wall-clock numbers taken during a sweep are untrustworthy.** Use load-independent evidence.\n\n## Clean up when you are done\n\nA sweep and its debugging leave several distinct piles. Delete all of them:\n\n- the harness's own fixture roots (`mkdtemp` dirs under the cache);\n- any ad-hoc reproduction trees you created while triaging — these are the biggest surprise,\n  measured at **12 GB** from one night of hand-testing;\n- temp catalogs and worklists in `\u002Ftmp`;\n- per-cell logs for records you have finished analyzing, if the verdict is recorded.\n\nKeep the `results.json` records — they are small (~39 KB each, ~0.1 GB for a 2,250 corpus) and they\nare the actual output.\n\n## Stopping to fix the harness\n\nEvery harness fix invalidates the records taken before it, because a record means something\ndifferent under a changed instrument. That is what provenance hashes exist to expose.\n\n- **Purge exactly the records the fix could change**, not all of them, when you can characterise the\n  set (e.g. \"only runs pinned to Node ≤15\"). Purge everything when the change is corpus-wide.\n- **Never edit the harness while a batch runs.** The driver script is re-read on every package\n  spawn; editing it mid-run silently corrupted 54 of 100 packages once. Stop, fix, restart.\n- A binary rebuild mid-run is safe IF the runner snapshots the binary at startup. Confirm it does.\n- Weigh the trade honestly: a correct instrument with partial coverage beats a complete corpus\n  measured by an instrument you already know is wrong. But say the coverage number plainly.\n\n## Bringing the harness up on a NEW PLATFORM\n\n**Load the `probe-platforms` skill.** It owns the bring-up ladder (debug with the SMALLEST payload —\nsix Windows faults were found using a 25-minute package when seconds would have done), the Windows\nspawn\u002Fpath\u002Fdisk faults, the Linux Landlock and node-layout traps, and the remote-shell mechanics.\n\nTwo rules from it that bite during a sweep specifically:\n\n- **Shuffle the worklist before sharding.** It is name-sorted, so a contiguous slice hands every\n  shard the same heavy family at once — four shards once sat 12 minutes on one family and produced\n  zero records. A seeded shuffle fixed it in seconds.\n- **Confirm the override actually ENGAGED before believing any cell.** The catalog parser rejects a\n  malformed catalog (e.g. `read` alongside `write: \"disk\"`, which is redundant) and falls back to the\n  compiled-in one *silently* — so a run you believe grants network may have none.\n\n## Reading results\n\n- **Coverage first, intersected with the worklist.** Counting every record on disk inflates it with\n  earlier runs. Packages with NO record mean the recorded set is a biased sample — the heavy native\n  builds are exactly the ones that fail, so survivors are not the corpus.\n- **Read the FIRST error, never the tail.** These logs end in a stack trace and a summary; the cause\n  is ~40 lines earlier.\n- **Group by CAUSE before concluding.** A cluster of similar-looking strings is not evidence of a\n  shared cause; clustering by an error substring once merged two unrelated bugs.\n- **Frequency tables must count DISTINCT PACKAGES, not records.** One package measured at four\n  versions looks like ecosystem-wide leakage otherwise.\n",{},"main","9b734566fd36ceed8379f67a92291bc68af60801",".claude\u002Fskills\u002Fcorpus-sweep",71,"https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub","https:\u002F\u002Fnubjs.com","https:\u002F\u002Fgithub.com\u002Fnubjs","The all-in-one Node.js toolkit","2026-11-14 14:11:35",{"source":7,"skillName":5,"commands":33,"prompt":43,"zipAvailable":44},[34,37,40],{"tool":35,"command":36},"npx","npx skills add nubjs\u002Fnub --skill corpus-sweep",{"tool":38,"command":39},"bunx","bunx skills add nubjs\u002Fnub --skill corpus-sweep",{"tool":41,"command":42},"pnpm","pnpm dlx skills add nubjs\u002Fnub --skill corpus-sweep","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fcorpus-sweep\u002FSKILL.md",true,[],[47,53,59,66,73,80],{"skillNo":48,"name":49,"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":50,"categories":51,"keywords":52,"publishedAt":18},"01M758BSH88RVXBG8TPCT427TY","nub","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002Fskills\u002Fnub\u002FSKILL.md",[],[],{"skillNo":54,"name":49,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M758BKPRHXS5MS27WG908PBG","This project uses nub, a single Rust CLI that runs on the project's installed Node and replaces the everyday Node toolchain. ALWAYS prefer nub over node, bun, npm, npx, pnpm, and yarn for: executing JS\u002FTS files, running package.json scripts, the npx-equivalent, installing dependencies, watch mode, and Node version management. Surfaces whenever you would otherwise reach for node\u002Fbun\u002Fnpm\u002Fnpx or a package manager.","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002Fsite\u002Fpublic\u002Fskill.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":62,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":63,"categories":64,"keywords":65,"publishedAt":18},"01M758BCR8A21H8P4TQ0YRHTC7","worktree","Create and manage isolated git worktrees for parallel build\u002Ftest\u002Flanding work on the nub repo. Invoke (via the Skill tool) whenever you need a fresh worktree to land a change, when you want to know what `.worktreeinclude` does or how to add an entry, or when cleaning up after a merge. Encodes the one-command setup (`nub scripts\u002Fnew-worktree.ts \u003Cslug>` or `node …`) that bakes in the proven recipe — worktree off origin\u002Fmain (vendor\u002Faube is plain in-tree files now, no submodule init), the shared CARGO_TARGET_DIR (`~\u002F.cache\u002Fnub\u002Fshared-target`) that all worktrees reuse, and applying `.worktreeinclude` — plus the eagerly-pull-the-shared-tree discipline and the safe cleanup path. Pairs with the `dev-loop` build skill.","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fworktree\u002FSKILL.md",[],[],{"skillNo":67,"name":68,"nameZh":6,"description":69,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":70,"categories":71,"keywords":72,"publishedAt":18},"01M758B9NZ1GQMD1A6B25NK0D9","windows-vm-test","Run ad-hoc Nub tests and debugging probes in the maintainer's existing local Windows 11 ARM64 QEMU\u002FHVF VM when its external disk and runbook are present. Use for interactive win32-arm64 or Windows-on-ARM x64-emulation checks; use ci-adhoc-test for true x64-native behavior or when the local VM is unavailable.","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fwindows-vm-test\u002FSKILL.md",[],[],{"skillNo":74,"name":75,"nameZh":6,"description":76,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":77,"categories":78,"keywords":79,"publishedAt":18},"01M758B2JHMBB6YJTGYSJR65S4","visual-review","Verify UI\u002Flayout\u002Fstyling changes are correct by computing occlusion, clipping, and alignment from the browser's resolved paint order via the chrome-devtools MCP `evaluate_script` tool — instead of eyeballing a flat screenshot. Invoke BEFORE declaring any UI, site, or styling\u002Flayout change correct. Screenshots have no depth buffer, so z-index\u002Focclusion\u002Fclip bugs are exactly where \"just look at it\" fails; the `evaluate_script` routines below turn those fuzzy visual judgments into deterministic measurements — including optical center-of-mass, which measures the glyph ink's true visual center so differently-sized labels can be aligned by more than eye.","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fvisual-review\u002FSKILL.md",[],[],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":7,"ownerLogin":8,"ownerAvatarUrl":9,"ownerAccountNo":10,"ownerDisplayName":11,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"publishedAt":18},"01M758AWPM7CE440NBAYJHD44J","type-declarations","Add, audit, or update @nubjs\u002Ftypes declarations for Nub runtime APIs. Use whenever Nub adds or changes a user-visible global, built-in method, module loader type, or TC39 proposal; when TypeScript or @types\u002Fnode gains an overlapping declaration; and before every release to reconcile runtime changes since the previous tag. Covers conflict-safe global augmentation, compiler-version routing, fixture matrices, and package verification.","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Ftype-declarations\u002FSKILL.md",[],[]]