[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75898XHVVSKRB02ABRH5RDW":3,"skill-same:01M75898XHVVSKRB02ABRH5RDW":47},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":29,"ownerHtmlUrl":30,"ownerBlogUrl":29,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M75898XHVVSKRB02ABRH5RDW","probe-platforms",null,"Bring a test harness or probe up on a NEW operating system (Windows or Linux), or debug one that fails there. Invoke before the first run on a platform, and whenever a harness works on one OS but not another. Carries the bring-up ladder that avoids spending hours learning single booleans, the Windows spawn\u002Fpath\u002Fdisk faults that each present as something else entirely, the Linux Landlock and node-layout traps, and the remote-shell mechanics (PowerShell-over-SSH quoting, pkill matching your own command) that waste a loop each time they are rediscovered.","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\u002Fprobe-platforms\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: probe-platforms\ndescription: Bring a test harness or probe up on a NEW operating system (Windows or Linux), or debug one that fails there. Invoke before the first run on a platform, and whenever a harness works on one OS but not another. Carries the bring-up ladder that avoids spending hours learning single booleans, the Windows spawn\u002Fpath\u002Fdisk faults that each present as something else entirely, the Linux Landlock and node-layout traps, and the remote-shell mechanics (PowerShell-over-SSH quoting, pkill matching your own command) that waste a loop each time they are rediscovered.\n---\n\n# Bringing a probe up on a new platform\n\nA harness that works on one OS fails on another for reasons that are almost never the thing the\nerror names. Expect a CHAIN of faults, each hidden by the one in front of it.\n\n## ⛔ Debug with the SMALLEST payload — the single most expensive mistake here\n\nMeasured, bringing the build-jail probe up on Windows: **six sequential faults, found using\n`puppeteer` as the probe, at ~25 minutes per attempt.** Every one was answerable by a question as\nsmall as *\"does the binary spawn?\"* — seconds with `is-odd`. About two hours bought six booleans.\n\n**Climb this ladder in order. Do not skip to the end.**\n\n| # | Step | Proves |\n|---|---|---|\n| 1 | `\u003Cbinary> --version` | it exists, is executable, spawns |\n| 2 | same, with the config\u002Foverride env set | the feature is compiled in AND the override *engages* |\n| 3 | install a trivial no-scripts package (`is-odd`) | fixture, store, linker |\n| 4 | one tiny package that DOES run a script | a child process spawns under confinement |\n| 5 | the real canary (large, real download) | end-to-end — a FINAL GATE, never a debugging instrument |\n\nEach step is seconds. Each isolates one layer. **When one passes, expect the next to fail** — do not\nread a green step as \"the platform works.\"\n\n**The top fault usually accuses the wrong subsystem.** On Windows the probe insisted the binary\nlacked a cargo feature; the truth four layers down was an unspawnable path, with every per-cell log\nat 0 bytes. When a diagnostic names a subsystem, verify that subsystem independently before acting\non the accusation — and when you fix a misleading diagnostic, it pays for itself immediately.\n\n## Windows — all measured\n\n- **`for \u002Ff` runs its command through `cmd \u002Fc`, which STRIPS the outer quote pair** when the string\n  both starts and ends with a quote. So the natural spelling of a capture,\n  `` for \u002Ff \"usebackq delims=\" %%i in (`\"%EXE%\" arg \"%DIR%\"`) ``, degrades to\n  `C:\\...\\x.exe\" arg \"C:\\...\\dir` and dies with **\"The filename, directory name, or volume label\n  syntax is incorrect\"** — with or without spaces in the path. Fix: wrap the whole command in one\n  MORE quote pair (`` `\"\"%EXE%\" arg \"%DIR%\"\"` ``). Measured against a temp-file-redirect alternative,\n  which did *not* work.\n- **An undefined `%VAR%` expands to its own literal text**, so a missing variable is passed onward as\n  the string `%VAR%` rather than being empty — silently handing a bogus path downstream. Guard with\n  `if not defined VAR set \"VAR=%CD%\"`, under `setlocal` so the default does not leak to the caller.\n- **A `.cmd` on `PATH` is invisible to whole classes of CI.** In this repo `native-deps.yml` is the\n  only workflow touching node-gyp and it is ubuntu-only, so the Windows `.cmd` shim shipped broken and\n  unnoticed for as long as another code path kept it from ever running. Green Windows jobs are not\n  evidence for a Windows file nothing executes — check that some job actually runs it.\n- **Node aborts at startup inside nub's build jail on Windows**: `Assertion failed:\n  ncrypto::CSPRNG(nullptr, 0)`. Any jailed scenario whose script is `node` therefore cannot report on\n  what it meant to test; skip it with the reason printed rather than recording a pass or a fail.\n\n- **`spawnSync` cannot run `npm`\u002F`npx`\u002F`pnpm`.** They are `.cmd` shims: the bare name gives\n  **ENOENT**, the `.cmd` spelling gives **EINVAL** (Node has refused to `CreateProcess` a batch file\n  since CVE-2024-27980). `shell: true` works but is DEP0190 — args are concatenated rather than\n  escaped, so cmd.exe re-parses a spec like `@scope\u002Fpkg@1.0.0`. **Run the bundled JS instead:\n  `node \u003Cpath>\u002Fnode_modules\u002Fnpm\u002Fbin\u002Fnpm-cli.js`.**\n- **Git Bash paths are not spawnable by native tools.** `cd $(dirname) && pwd` under Git Bash yields\n  `\u002Fc\u002FUsers\u002F…`, handed to CreateProcess verbatim → ENOENT. Convert to `C:\\…`; for any path passed to\n  a Windows binary use `cygpath -w`. `\u002Ftmp\u002Fx` is NOT where a Windows `node` will look.\n- **A copied binary must keep its `.exe`.** Windows decides executability from the SUFFIX, so a\n  content-addressed copy named by bare hash is ENOENT even at 1.2 GB and mode 0755.\n- **Removing a `node_modules` tree fails with EPERM routinely** — a lifecycle child still holds a\n  handle, or an indexer\u002FAV opens files behind you. Use `maxRetries`\u002F`retryDelay`, and **never let a\n  cleanup failure discard a measurement that already succeeded** (a non-zero exit from tidy-up reads\n  upstream as \"no result\").\n- **Disk is a first-class hazard, and a full disk is SILENT AND EVIL.** A debug Rust binary can be\n  ~1.2 GB with a `target\u002F` of tens of GB; add leaked fixture trees and 80 GB vanishes. When it\n  fills, `git fetch` fails and the probe runs **STALE CODE while you believe you are testing the\n  fix**. Size 200 GB+, build `--profile fast` not `debug`, and check free space before trusting any\n  run. Growing the GCE disk is not enough — the in-OS partition must be extended too, and\n  `Resize-Partition` can simply fail.\n- **Prefer a CI-built binary to provisioning MSVC.** A `windows-latest` workflow that uploads\n  `target\u002Fdebug\u002Fnub.exe` gives a real MSVC binary for one dispatch, versus installing the VS\n  `VCTools` workload, `cmake`, and switching the rustup host triple by hand. A `windows-gnu`\n  cross-compile is NOT a substitute when the thing under test is OS confinement behaviour.\n- If you do build on the VM: rustup may default to `windows-gnu` (nub needs `-msvc`), VS Build Tools\n  can be present with **no C++ workload** so there is no linker at all, and `\u002FSTACK:8388608` is an\n  **MSVC-only** flag that breaks a GNU link — it is required under MSVC because Windows gives the\n  main thread 1 MB against Linux's 8 MB.\n\n### Windows filesystem speed — budget ~6x Linux, and beware the benchmark itself\n\nMEASURED on matched GCE `e2-standard-8` \u002F `pd-balanced` VMs, 3,000 small file creates:\n\n| | time |\n|---|---|\n| Linux (bash redirect) | **181 ms** |\n| Windows (`[System.IO.File]::WriteAllText`) | **1,142 ms** |\n| Windows (PowerShell `Set-Content`) | 3,156 ms |\n\n**~6x, not 17x.** The first Windows figure was `Set-Content`, whose cmdlet\u002Fpipeline overhead is\nroughly two-thirds of it — comparing that against a bash redirect is not varying one thing. Use a\nraw write on both sides before quoting any cross-OS I\u002FO ratio.\n\n⛔ **DO NOT EXTRAPOLATE THAT RATIO TO INSTALL TIME — I did, and it was wrong.** A microbenchmark of\nraw file creation is not a workload. Measured with the SAME nub binary on matched VMs:\n\n| install | Linux | Windows | ratio |\n|---|---|---|---|\n| trivial (`is-odd`) | 493 ms | 744 ms | **1.5x** |\n| file-heavier (`typescript`) | 1,675 ms | 3,218 ms | **1.9x** |\n\nSo a real install is **~1.5–2x**, not 6x: installs are dominated by network and archive work, and\nthe file-create penalty is a minority of the total. I first wrote \"budget an order of magnitude\" here\noff the microbenchmark alone; that was a wrong planning number in a durable doc. **Time the actual\nworkload before sizing anything.**\n\n(Note the file COUNTS in that test are not comparable — `Get-ChildItem -Recurse` reported 9 where\n`find -L` reported 264, because neither traverses junctions the same way. The TIMES are the\ncomparable part; see the junction-traversal trap above.)\n\n## Linux — all measured\n\n- **An AUTHORED Landlock grant naming a path that does not exist makes the whole policy\n  uncompilable** — refused, not degraded (`PolicyNotExpressible`). Speculative grants skip a missing\n  path; authored ones abort. One bad entry breaks every confined run.\n- **Do not read \"Landlock unavailable\" as a kernel problem.** That error covers both an old kernel\n  and *our own policy failing to compile*; it claimed a missing feature on a 6.17 kernel with ABI 4.\n- **A fresh cloud box has no system `node`**, and the Rust build needs `cmake`. Provision Nodes with\n  the tool under test where possible (dogfoods the real mechanism) — but its layout may be\n  `\u003Ccache>\u002Fnub\u002Fnode\u002F22.23.1\u002Fbin`, with **no `v` prefix**, where nvm uses `v22.23.1`. Anything parsing\n  a version out of a node path must accept both or it silently returns null.\n- **Absent tooling is environmental, not a defect.** `pnpm` missing → a script shelling out to it\n  exits 127; a `-musl` package on a glibc box cannot load. Neither is a bug in the thing under test.\n- **Old Node pins bring old npm, and npm 6 races on its own `_cacache` under concurrency** —\n  surfacing as `rimraf: missing path` + `Callback called more than once`. This manufactures FALSE\n  defect verdicts that a double-control cannot catch, because both attempts sit in the same busy\n  window. Re-verify any defect verdict serially once the batch drains.\n\n## Remote-shell mechanics that cost a loop every time\n\n- **SSH to Windows lands in PowerShell, not bash.** `bash.exe -lc \"…\"` must go through a `.ps1` or\n  PowerShell parses `-lc` as an expression. Nested quoting through\n  `ssh … powershell -Command \"…\"` breaks constantly — prefer a command containing **no double\n  quotes at all**, or `scp` a script file (which fails on a full disk, so check that first).\n- **Keep every PowerShell script ASCII.** One em-dash anywhere fails with `The string is missing the\n  terminator`, pointing at the LAST line of the file rather than the offending one.\n- **`pkill -f \u003Cpattern>` matches YOUR OWN command line.** Killing `-f \"search.mjs\"` from a shell\n  whose command text contains `search.mjs` kills that shell — the remote command dies mid-way and\n  returns no output, which reads like a hang. Split the kill into its own invocation, or break the\n  literal (`\"sea\"\"rch.mjs\"`).\n- **Detach long remote runs with `tmux new-session -d`.** A bare `nohup … &` over SSH hangs the\n  connection and can leave only the first job running.\n- **Prefer native tools for bulk filesystem work.** `du -sh` over Git Bash on a large Windows tree\n  is glacial and will time out; PowerShell equivalents finish.\n","# Bringing a probe up on a new platform\n\nA harness that works on one OS fails on another for reasons that are almost never the thing the\nerror names. Expect a CHAIN of faults, each hidden by the one in front of it.\n\n## ⛔ Debug with the SMALLEST payload — the single most expensive mistake here\n\nMeasured, bringing the build-jail probe up on Windows: **six sequential faults, found using\n`puppeteer` as the probe, at ~25 minutes per attempt.** Every one was answerable by a question as\nsmall as *\"does the binary spawn?\"* — seconds with `is-odd`. About two hours bought six booleans.\n\n**Climb this ladder in order. Do not skip to the end.**\n\n| # | Step | Proves |\n|---|---|---|\n| 1 | `\u003Cbinary> --version` | it exists, is executable, spawns |\n| 2 | same, with the config\u002Foverride env set | the feature is compiled in AND the override *engages* |\n| 3 | install a trivial no-scripts package (`is-odd`) | fixture, store, linker |\n| 4 | one tiny package that DOES run a script | a child process spawns under confinement |\n| 5 | the real canary (large, real download) | end-to-end — a FINAL GATE, never a debugging instrument |\n\nEach step is seconds. Each isolates one layer. **When one passes, expect the next to fail** — do not\nread a green step as \"the platform works.\"\n\n**The top fault usually accuses the wrong subsystem.** On Windows the probe insisted the binary\nlacked a cargo feature; the truth four layers down was an unspawnable path, with every per-cell log\nat 0 bytes. When a diagnostic names a subsystem, verify that subsystem independently before acting\non the accusation — and when you fix a misleading diagnostic, it pays for itself immediately.\n\n## Windows — all measured\n\n- **`for \u002Ff` runs its command through `cmd \u002Fc`, which STRIPS the outer quote pair** when the string\n  both starts and ends with a quote. So the natural spelling of a capture,\n  `` for \u002Ff \"usebackq delims=\" %%i in (`\"%EXE%\" arg \"%DIR%\"`) ``, degrades to\n  `C:\\...\\x.exe\" arg \"C:\\...\\dir` and dies with **\"The filename, directory name, or volume label\n  syntax is incorrect\"** — with or without spaces in the path. Fix: wrap the whole command in one\n  MORE quote pair (`` `\"\"%EXE%\" arg \"%DIR%\"\"` ``). Measured against a temp-file-redirect alternative,\n  which did *not* work.\n- **An undefined `%VAR%` expands to its own literal text**, so a missing variable is passed onward as\n  the string `%VAR%` rather than being empty — silently handing a bogus path downstream. Guard with\n  `if not defined VAR set \"VAR=%CD%\"`, under `setlocal` so the default does not leak to the caller.\n- **A `.cmd` on `PATH` is invisible to whole classes of CI.** In this repo `native-deps.yml` is the\n  only workflow touching node-gyp and it is ubuntu-only, so the Windows `.cmd` shim shipped broken and\n  unnoticed for as long as another code path kept it from ever running. Green Windows jobs are not\n  evidence for a Windows file nothing executes — check that some job actually runs it.\n- **Node aborts at startup inside nub's build jail on Windows**: `Assertion failed:\n  ncrypto::CSPRNG(nullptr, 0)`. Any jailed scenario whose script is `node` therefore cannot report on\n  what it meant to test; skip it with the reason printed rather than recording a pass or a fail.\n\n- **`spawnSync` cannot run `npm`\u002F`npx`\u002F`pnpm`.** They are `.cmd` shims: the bare name gives\n  **ENOENT**, the `.cmd` spelling gives **EINVAL** (Node has refused to `CreateProcess` a batch file\n  since CVE-2024-27980). `shell: true` works but is DEP0190 — args are concatenated rather than\n  escaped, so cmd.exe re-parses a spec like `@scope\u002Fpkg@1.0.0`. **Run the bundled JS instead:\n  `node \u003Cpath>\u002Fnode_modules\u002Fnpm\u002Fbin\u002Fnpm-cli.js`.**\n- **Git Bash paths are not spawnable by native tools.** `cd $(dirname) && pwd` under Git Bash yields\n  `\u002Fc\u002FUsers\u002F…`, handed to CreateProcess verbatim → ENOENT. Convert to `C:\\…`; for any path passed to\n  a Windows binary use `cygpath -w`. `\u002Ftmp\u002Fx` is NOT where a Windows `node` will look.\n- **A copied binary must keep its `.exe`.** Windows decides executability from the SUFFIX, so a\n  content-addressed copy named by bare hash is ENOENT even at 1.2 GB and mode 0755.\n- **Removing a `node_modules` tree fails with EPERM routinely** — a lifecycle child still holds a\n  handle, or an indexer\u002FAV opens files behind you. Use `maxRetries`\u002F`retryDelay`, and **never let a\n  cleanup failure discard a measurement that already succeeded** (a non-zero exit from tidy-up reads\n  upstream as \"no result\").\n- **Disk is a first-class hazard, and a full disk is SILENT AND EVIL.** A debug Rust binary can be\n  ~1.2 GB with a `target\u002F` of tens of GB; add leaked fixture trees and 80 GB vanishes. When it\n  fills, `git fetch` fails and the probe runs **STALE CODE while you believe you are testing the\n  fix**. Size 200 GB+, build `--profile fast` not `debug`, and check free space before trusting any\n  run. Growing the GCE disk is not enough — the in-OS partition must be extended too, and\n  `Resize-Partition` can simply fail.\n- **Prefer a CI-built binary to provisioning MSVC.** A `windows-latest` workflow that uploads\n  `target\u002Fdebug\u002Fnub.exe` gives a real MSVC binary for one dispatch, versus installing the VS\n  `VCTools` workload, `cmake`, and switching the rustup host triple by hand. A `windows-gnu`\n  cross-compile is NOT a substitute when the thing under test is OS confinement behaviour.\n- If you do build on the VM: rustup may default to `windows-gnu` (nub needs `-msvc`), VS Build Tools\n  can be present with **no C++ workload** so there is no linker at all, and `\u002FSTACK:8388608` is an\n  **MSVC-only** flag that breaks a GNU link — it is required under MSVC because Windows gives the\n  main thread 1 MB against Linux's 8 MB.\n\n### Windows filesystem speed — budget ~6x Linux, and beware the benchmark itself\n\nMEASURED on matched GCE `e2-standard-8` \u002F `pd-balanced` VMs, 3,000 small file creates:\n\n| | time |\n|---|---|\n| Linux (bash redirect) | **181 ms** |\n| Windows (`[System.IO.File]::WriteAllText`) | **1,142 ms** |\n| Windows (PowerShell `Set-Content`) | 3,156 ms |\n\n**~6x, not 17x.** The first Windows figure was `Set-Content`, whose cmdlet\u002Fpipeline overhead is\nroughly two-thirds of it — comparing that against a bash redirect is not varying one thing. Use a\nraw write on both sides before quoting any cross-OS I\u002FO ratio.\n\n⛔ **DO NOT EXTRAPOLATE THAT RATIO TO INSTALL TIME — I did, and it was wrong.** A microbenchmark of\nraw file creation is not a workload. Measured with the SAME nub binary on matched VMs:\n\n| install | Linux | Windows | ratio |\n|---|---|---|---|\n| trivial (`is-odd`) | 493 ms | 744 ms | **1.5x** |\n| file-heavier (`typescript`) | 1,675 ms | 3,218 ms | **1.9x** |\n\nSo a real install is **~1.5–2x**, not 6x: installs are dominated by network and archive work, and\nthe file-create penalty is a minority of the total. I first wrote \"budget an order of magnitude\" here\noff the microbenchmark alone; that was a wrong planning number in a durable doc. **Time the actual\nworkload before sizing anything.**\n\n(Note the file COUNTS in that test are not comparable — `Get-ChildItem -Recurse` reported 9 where\n`find -L` reported 264, because neither traverses junctions the same way. The TIMES are the\ncomparable part; see the junction-traversal trap above.)\n\n## Linux — all measured\n\n- **An AUTHORED Landlock grant naming a path that does not exist makes the whole policy\n  uncompilable** — refused, not degraded (`PolicyNotExpressible`). Speculative grants skip a missing\n  path; authored ones abort. One bad entry breaks every confined run.\n- **Do not read \"Landlock unavailable\" as a kernel problem.** That error covers both an old kernel\n  and *our own policy failing to compile*; it claimed a missing feature on a 6.17 kernel with ABI 4.\n- **A fresh cloud box has no system `node`**, and the Rust build needs `cmake`. Provision Nodes with\n  the tool under test where possible (dogfoods the real mechanism) — but its layout may be\n  `\u003Ccache>\u002Fnub\u002Fnode\u002F22.23.1\u002Fbin`, with **no `v` prefix**, where nvm uses `v22.23.1`. Anything parsing\n  a version out of a node path must accept both or it silently returns null.\n- **Absent tooling is environmental, not a defect.** `pnpm` missing → a script shelling out to it\n  exits 127; a `-musl` package on a glibc box cannot load. Neither is a bug in the thing under test.\n- **Old Node pins bring old npm, and npm 6 races on its own `_cacache` under concurrency** —\n  surfacing as `rimraf: missing path` + `Callback called more than once`. This manufactures FALSE\n  defect verdicts that a double-control cannot catch, because both attempts sit in the same busy\n  window. Re-verify any defect verdict serially once the batch drains.\n\n## Remote-shell mechanics that cost a loop every time\n\n- **SSH to Windows lands in PowerShell, not bash.** `bash.exe -lc \"…\"` must go through a `.ps1` or\n  PowerShell parses `-lc` as an expression. Nested quoting through\n  `ssh … powershell -Command \"…\"` breaks constantly — prefer a command containing **no double\n  quotes at all**, or `scp` a script file (which fails on a full disk, so check that first).\n- **Keep every PowerShell script ASCII.** One em-dash anywhere fails with `The string is missing the\n  terminator`, pointing at the LAST line of the file rather than the offending one.\n- **`pkill -f \u003Cpattern>` matches YOUR OWN command line.** Killing `-f \"search.mjs\"` from a shell\n  whose command text contains `search.mjs` kills that shell — the remote command dies mid-way and\n  returns no output, which reads like a hang. Split the kill into its own invocation, or break the\n  literal (`\"sea\"\"rch.mjs\"`).\n- **Detach long remote runs with `tmux new-session -d`.** A bare `nohup … &` over SSH hangs the\n  connection and can leave only the first job running.\n- **Prefer native tools for bulk filesystem work.** `du -sh` over Git Bash on a large Windows tree\n  is glacial and will time out; PowerShell equivalents finish.\n",{"name":5,"description":7},"main","9b734566fd36ceed8379f67a92291bc68af60801",".claude\u002Fskills\u002Fprobe-platforms",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":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add nubjs\u002Fnub --skill probe-platforms",{"tool":39,"command":40},"bunx","bunx skills add nubjs\u002Fnub --skill probe-platforms",{"tool":42,"command":43},"pnpm","pnpm dlx skills add nubjs\u002Fnub --skill probe-platforms","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fprobe-platforms\u002FSKILL.md",true,[],[48,54,60,67,74,81],{"skillNo":49,"name":50,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":51,"categories":52,"keywords":53,"publishedAt":19},"01M758BSH88RVXBG8TPCT427TY","nub","https:\u002F\u002Fgithub.com\u002Fnubjs\u002Fnub\u002Fblob\u002Fmain\u002Fskills\u002Fnub\u002FSKILL.md",[],[],{"skillNo":55,"name":50,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"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":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"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":68,"name":69,"nameZh":6,"description":70,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":71,"categories":72,"keywords":73,"publishedAt":19},"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":75,"name":76,"nameZh":6,"description":77,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":78,"categories":79,"keywords":80,"publishedAt":19},"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":82,"name":83,"nameZh":6,"description":84,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":85,"categories":86,"keywords":87,"publishedAt":19},"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",[],[]]