[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75JPCF543BA5647DJNKBJDB":3,"skill-same:01M75JPCF543BA5647DJNKBJDB":70},{"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":31,"ownerDescription":32,"updatedAt":33,"install":34,"resources":47},"01M75JPCF543BA5647DJNKBJDB","codespine-workload",null,"Benchmark a program in a controlled environment and attribute the cost with the code graph. Use when asked to profile under a CPU\u002Fmemory cap, measure server capacity (\"how many requests per second before p99\u002Flatency breaks an SLO\"), decide \"when do I need more servers\", find where runtime is spent under load, or run a workload in a constrained \"one box\" container. Two modes: cpu-profile (where the time goes) and loadtest (how much load it takes). Requires the codespine CLI; Docker is optional (falls back to running on the host).","jeromeetienne\u002Fcodespine","jeromeetienne","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F252962?v=4","01M75JP4GNBAWHWK7J2DEKE33H","Jerome Etienne","01M75JP4HTYYDYGRM1QN5G8816",7,"MIT","https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-workload\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: codespine-workload\ndescription: >-\n  Benchmark a program in a controlled environment and attribute the cost with the\n  code graph. Use when asked to profile under a CPU\u002Fmemory cap, measure server\n  capacity (\"how many requests per second before p99\u002Flatency breaks an SLO\"),\n  decide \"when do I need more servers\", find where runtime is spent under load, or\n  run a workload in a constrained \"one box\" container. Two modes: cpu-profile\n  (where the time goes) and loadtest (how much load it takes). Requires the\n  codespine CLI; Docker is optional (falls back to running on the host).\n---\n\n# codespine-workload\n\nExercise a program under a chosen environment, then attribute the cost back onto\nthe code graph. This turns vague runtime questions (\"why is it slow?\", \"how many\nrequests can it take?\", \"when do I add a server?\") into measured, graph-grounded\nanswers.\n\nThe capability is one idea in three parts — **workload kind × environment ×\nattribution**:\n\n- **kind** — `cpu-profile` (loop the hot path) *or* `loadtest` (ramp a running server).\n- **environment** — `--host` (uncapped baseline) *or* a container under an enforced\n  `--cpus`\u002F`--memory` cap (a realistic \"one server\" box).\n- **attribution** — `cpu-profile` → `enrich` → `hotspots`\u002F`cost` (graph-grounded\n  *where the time goes*); `loadtest` → a latency-vs-throughput curve and the knee\n  (*how many requests\u002Fsecond before the SLO breaks*).\n\n## When to use this skill\n\n- \"Where is the time spent \u002F why is this slow under load?\" → **cpu-profile**.\n- \"How many requests\u002Fsecond can my server handle before p99 crosses my SLO?\" → **loadtest**.\n- \"When do I need another server \u002F how do I size capacity?\" → **loadtest**.\n- \"Will it hold up on a small box (0.5 CPU \u002F 512 MB)?\" → either mode, **docker** environment.\n- The agent is mid-task in `\u002Fcodespine-interview` or `\u002Fcodespine-optimize` and hits a\n  `needs-workload` situation (it must *measure* a runtime claim, not guess) → use this.\n\nFor deciding *which symbol* matters from an existing profile, use the\n`codespine-query` skill (`hotspots`, `cost`). This skill is about *producing* the\nmeasurement in a controlled environment in the first place.\n\n## Realism vs. determinism — the honesty rule\n\nThere are two tracks, and they answer different questions. Keep them separate:\n\n- **Determinism track** (the `benchmark` command): low-variance before\u002Fafter deltas\n  for \"did my edit make X faster?\". Stable median, advisory.\n- **Realism track** (this skill, under a cap): \"will it hold up on a constrained\n  box?\". It intentionally introduces a cap and **accepts scheduler noise**. These\n  numbers are a capacity *estimate*, **not** a deterministic gate. Never diff two\n  realism runs as a pass\u002Ffail, and never use a realism number to approve an\n  optimization — that is what `benchmark`\u002F`verify` are for.\n\nState this caveat whenever you report realism-track numbers. See\n[`reference\u002Ftwo-track.md`](reference\u002Ftwo-track.md) for the full framing.\n\n## Prerequisites\n\n- The **codespine CLI** (`codespine …`, or `npx codespine …`).\n- For **cpu-profile attribution**: a built graph at `.\u002F.codespine\u002Fgraph.kuzu`. If it\n  is missing, build it once: `codespine extract . --semantic` then `codespine load`.\n- For the **docker** environment: a container runtime (Docker Desktop \u002F OrbStack \u002F\n  podman). If absent, use `--host`.\n- For **loadtest**: the server must be runnable locally and expose a readiness\n  endpoint (something that returns 2xx when ready).\n\n## Where artifacts live, and what this skill ships\n\nAll per-project artifacts go in the **user's** repo under `.\u002F.codespine\u002Fworkload\u002F`\n(gitignorable) — never in the codespine package. This skill ships starting-point\ntemplates next to this file under [`templates\u002F`](templates\u002F):\n\n- [`templates\u002FDockerfile`](templates\u002FDockerfile) — a tiny `node:24-slim` runner image (`tsx` + `autocannon`).\n- [`templates\u002Fcpu_profile_driver.template.ts`](templates\u002Fcpu_profile_driver.template.ts) — skeleton for the cpu-profile loop.\n- [`templates\u002Floadtest_driver.template.ts`](templates\u002Floadtest_driver.template.ts) — a working HTTP capacity driver; edit the marked block for your endpoints.\n\nRun `codespine workload scaffold --kind cpu-profile|loadtest|both` to copy the\nrelevant template(s) into `.\u002F.codespine\u002Fworkload\u002F`, then fill in the\n`\u002F\u002F EDIT FOR YOUR PROJECT` block. The rest is generic.\n\n## Mode A — cpu-profile: where does the time go?\n\n1. **Build the graph** if needed: `codespine extract . --semantic && codespine load`.\n2. **Scaffold**: `codespine workload scaffold --kind cpu-profile` writes the driver +\n   Dockerfile to `.\u002F.codespine\u002Fworkload\u002F`.\n3. **Author the driver**: in the EDIT block of\n   `.\u002F.codespine\u002Fworkload\u002Fcpu_profile_driver.ts`, import the project's hot module(s)\n   and exercise the real public API in a deterministic loop (no timing, just work).\n   It must live *outside* the extracted source root so it never becomes a graph\n   node; imports are module-relative.\n4. **Run it** — one command profiles, enriches the graph, and prints the top self\n   time. On the **host** (simplest; uses the project's existing `node_modules`):\n   ```bash\n   codespine workload run --driver .\u002F.codespine\u002Fworkload\u002Fcpu_profile_driver.ts --root .\n   ```\n   …or under an enforced **cap** for the \"constrained box\" answer — `--project` is the\n   directory mounted at `\u002Fwork`, and the driver must live inside it:\n   ```bash\n   codespine workload run --driver .\u002F.codespine\u002Fworkload\u002Fcpu_profile_driver.ts --root . \\\n     --docker --cpus 0.5 --memory 512m --project .\n   ```\n   `--root` must be the directory you passed to `extract` (usually `.`, or a\n   subdirectory like `.\u002Fsrc` if you extracted one) — otherwise the profile's frame\n   paths will not resolve onto the graph and most samples come back \"unattributed\".\n5. **Read the ranked views**: `codespine hotspots --by self-time` (the leaves where\n   CPU burns) and `codespine cost` (which symbols the time is spent *under*). Report\n   the top frames, tie them to source, and — if asked — to optimization tasks.\n\n## Mode B — loadtest: how much load can it take?\n\n1. **Learn the server**: find the entrypoint (how it boots, what port), the\n   readiness endpoint, and the routes + their request contracts (read the route\n   handlers). For a write path, find the durability config (does it fsync?).\n2. **Scaffold + author**: copy `templates\u002Floadtest_driver.template.ts` to\n   `.\u002F.codespine\u002Fworkload\u002Floadtest_driver.ts` and edit the marked block:\n   how to start the server, the readiness path, and `buildRequests()` (your\n   endpoint mix). A filled example for a typical REST API is in the template's\n   comments.\n3. **Run** — host baseline first, then the cap:\n   ```bash\n   # host (uncapped upper bound):\n   SERVER_ENTRY=src\u002Fmain.ts PORT=3000 DB_PATH=\u002Ftmp\u002Fapp.loadtest.db \\\n     LOADTEST_PROFILE=mixed LOADTEST_SLO_P99_MS=200 \\\n     node --import tsx .codespine\u002Fworkload\u002Floadtest_driver.ts\n\n   # docker (one constrained \"server\" box — realism):\n   docker build -t codespine-workload-runner .codespine\u002Fworkload\n   docker run --rm --cpus 0.5 --memory 512m --memory-swap 512m \\\n     -v \"$PWD:\u002Fwork:ro\" -v \"$PWD\u002F.codespine\u002Fworkload:\u002Fopt\u002Frunner\u002Fwl\" \\\n     -e SERVER_ENTRY=\u002Fwork\u002Fsrc\u002Fmain.ts -e SERVER_CWD=\u002Fopt\u002Frunner \\\n     -e PORT=3000 -e DB_PATH=\u002Ftmp\u002Fapp.loadtest.db \\\n     -e LOADTEST_PROFILE=mixed -e LOADTEST_SLO_P99_MS=200 \\\n     codespine-workload-runner \\\n     node --import tsx \u002Fopt\u002Frunner\u002Fwl\u002Floadtest_driver.ts\n   ```\n4. **Read the ramp**: the driver prints a table (offered rps → p50\u002Fp99) and the\n   knee. **Capacity = the highest sustained rps with p99 under the SLO.** \"Time to\n   add a server\" = steady-state load approaching that number.\n5. **Critical**: use a **file-backed** datastore for any write-inclusive profile\n   (set `DB_PATH`\u002Fequivalent to a file). With an in-memory store, fsync is a no-op\n   and write capacity is fictitious.\n\n## Choosing host vs. docker\n\n- **Host** — uncapped, uses the project's own `node_modules`, no container friction.\n  The easy starting point and the *upper-bound* baseline. A single-threaded server\n  effectively uses ~one core regardless of host core count.\n- **Docker** — an enforced cgroup cap is the only way to get a faithful \"one\n  server\" box (macOS has no native per-process hard cap). Use it for the realism\n  answer. Co-locate the load client in the container against `127.0.0.1` (avoids the\n  host↔VM network-latency floor; the trade-off is the client shares the CPU cap).\n  If the project has runtime dependencies with **native addons**, the container\n  needs a *Linux* build of them (a macOS-arm64 `node_modules` will not run in the\n  Linux VM) — provision them into a volume or install in the image. When that is\n  painful, prefer host mode.\n\nRun both and compare: the gap tells you how much of the slowness is the cap (cgroup\nthrottling + a squeezed page cache) versus the program itself.\n\n## Interpreting and reporting\n\n- Always attach the **realism caveat** to capped numbers (noise; estimate not gate).\n- At low offered rates, **p99 is sample-thin** (few requests) — lengthen the step\n  duration for a stable tail, and say so.\n- A *sharp* knee (throughput plateaus while latency explodes) usually means a\n  **serial bottleneck** (e.g. a synchronous DB on one event loop) — more cores will\n  not help a single-threaded server; \"more servers\" means more processes\u002Freplicas.\n- Tie findings to the graph: `cpu-profile` → name the hot symbols from `hotspots`;\n  for a slow endpoint, read its handler → service → datastore path.\n\n## Limitations \u002F when to escalate\n\nThis MVP assumes the workload is runnable via `node`\u002F`tsx`. Projects needing a\nbuild step, a non-npm package manager, non-HTTP load, custom readiness, or auth\nneed the driver\u002Frun-command adapted by hand — note that to the user rather than\nreporting a bogus number. For cpu-profile, `codespine workload run` (host and\n`--docker`) and `codespine workload scaffold` already wrap this orchestration; the\n`loadtest` kind is run from the scaffolded driver directly for now. The cpu-profile\n`--docker` path auto-provisions a Linux build of the project's dependencies into a\ncached volume (keyed by the lockfile); the loadtest kind under `--docker` is not\nwired yet (use host).\n","# codespine-workload\n\nExercise a program under a chosen environment, then attribute the cost back onto\nthe code graph. This turns vague runtime questions (\"why is it slow?\", \"how many\nrequests can it take?\", \"when do I add a server?\") into measured, graph-grounded\nanswers.\n\nThe capability is one idea in three parts — **workload kind × environment ×\nattribution**:\n\n- **kind** — `cpu-profile` (loop the hot path) *or* `loadtest` (ramp a running server).\n- **environment** — `--host` (uncapped baseline) *or* a container under an enforced\n  `--cpus`\u002F`--memory` cap (a realistic \"one server\" box).\n- **attribution** — `cpu-profile` → `enrich` → `hotspots`\u002F`cost` (graph-grounded\n  *where the time goes*); `loadtest` → a latency-vs-throughput curve and the knee\n  (*how many requests\u002Fsecond before the SLO breaks*).\n\n## When to use this skill\n\n- \"Where is the time spent \u002F why is this slow under load?\" → **cpu-profile**.\n- \"How many requests\u002Fsecond can my server handle before p99 crosses my SLO?\" → **loadtest**.\n- \"When do I need another server \u002F how do I size capacity?\" → **loadtest**.\n- \"Will it hold up on a small box (0.5 CPU \u002F 512 MB)?\" → either mode, **docker** environment.\n- The agent is mid-task in `\u002Fcodespine-interview` or `\u002Fcodespine-optimize` and hits a\n  `needs-workload` situation (it must *measure* a runtime claim, not guess) → use this.\n\nFor deciding *which symbol* matters from an existing profile, use the\n`codespine-query` skill (`hotspots`, `cost`). This skill is about *producing* the\nmeasurement in a controlled environment in the first place.\n\n## Realism vs. determinism — the honesty rule\n\nThere are two tracks, and they answer different questions. Keep them separate:\n\n- **Determinism track** (the `benchmark` command): low-variance before\u002Fafter deltas\n  for \"did my edit make X faster?\". Stable median, advisory.\n- **Realism track** (this skill, under a cap): \"will it hold up on a constrained\n  box?\". It intentionally introduces a cap and **accepts scheduler noise**. These\n  numbers are a capacity *estimate*, **not** a deterministic gate. Never diff two\n  realism runs as a pass\u002Ffail, and never use a realism number to approve an\n  optimization — that is what `benchmark`\u002F`verify` are for.\n\nState this caveat whenever you report realism-track numbers. See\n[`reference\u002Ftwo-track.md`](reference\u002Ftwo-track.md) for the full framing.\n\n## Prerequisites\n\n- The **codespine CLI** (`codespine …`, or `npx codespine …`).\n- For **cpu-profile attribution**: a built graph at `.\u002F.codespine\u002Fgraph.kuzu`. If it\n  is missing, build it once: `codespine extract . --semantic` then `codespine load`.\n- For the **docker** environment: a container runtime (Docker Desktop \u002F OrbStack \u002F\n  podman). If absent, use `--host`.\n- For **loadtest**: the server must be runnable locally and expose a readiness\n  endpoint (something that returns 2xx when ready).\n\n## Where artifacts live, and what this skill ships\n\nAll per-project artifacts go in the **user's** repo under `.\u002F.codespine\u002Fworkload\u002F`\n(gitignorable) — never in the codespine package. This skill ships starting-point\ntemplates next to this file under [`templates\u002F`](templates\u002F):\n\n- [`templates\u002FDockerfile`](templates\u002FDockerfile) — a tiny `node:24-slim` runner image (`tsx` + `autocannon`).\n- [`templates\u002Fcpu_profile_driver.template.ts`](templates\u002Fcpu_profile_driver.template.ts) — skeleton for the cpu-profile loop.\n- [`templates\u002Floadtest_driver.template.ts`](templates\u002Floadtest_driver.template.ts) — a working HTTP capacity driver; edit the marked block for your endpoints.\n\nRun `codespine workload scaffold --kind cpu-profile|loadtest|both` to copy the\nrelevant template(s) into `.\u002F.codespine\u002Fworkload\u002F`, then fill in the\n`\u002F\u002F EDIT FOR YOUR PROJECT` block. The rest is generic.\n\n## Mode A — cpu-profile: where does the time go?\n\n1. **Build the graph** if needed: `codespine extract . --semantic && codespine load`.\n2. **Scaffold**: `codespine workload scaffold --kind cpu-profile` writes the driver +\n   Dockerfile to `.\u002F.codespine\u002Fworkload\u002F`.\n3. **Author the driver**: in the EDIT block of\n   `.\u002F.codespine\u002Fworkload\u002Fcpu_profile_driver.ts`, import the project's hot module(s)\n   and exercise the real public API in a deterministic loop (no timing, just work).\n   It must live *outside* the extracted source root so it never becomes a graph\n   node; imports are module-relative.\n4. **Run it** — one command profiles, enriches the graph, and prints the top self\n   time. On the **host** (simplest; uses the project's existing `node_modules`):\n   ```bash\n   codespine workload run --driver .\u002F.codespine\u002Fworkload\u002Fcpu_profile_driver.ts --root .\n   ```\n   …or under an enforced **cap** for the \"constrained box\" answer — `--project` is the\n   directory mounted at `\u002Fwork`, and the driver must live inside it:\n   ```bash\n   codespine workload run --driver .\u002F.codespine\u002Fworkload\u002Fcpu_profile_driver.ts --root . \\\n     --docker --cpus 0.5 --memory 512m --project .\n   ```\n   `--root` must be the directory you passed to `extract` (usually `.`, or a\n   subdirectory like `.\u002Fsrc` if you extracted one) — otherwise the profile's frame\n   paths will not resolve onto the graph and most samples come back \"unattributed\".\n5. **Read the ranked views**: `codespine hotspots --by self-time` (the leaves where\n   CPU burns) and `codespine cost` (which symbols the time is spent *under*). Report\n   the top frames, tie them to source, and — if asked — to optimization tasks.\n\n## Mode B — loadtest: how much load can it take?\n\n1. **Learn the server**: find the entrypoint (how it boots, what port), the\n   readiness endpoint, and the routes + their request contracts (read the route\n   handlers). For a write path, find the durability config (does it fsync?).\n2. **Scaffold + author**: copy `templates\u002Floadtest_driver.template.ts` to\n   `.\u002F.codespine\u002Fworkload\u002Floadtest_driver.ts` and edit the marked block:\n   how to start the server, the readiness path, and `buildRequests()` (your\n   endpoint mix). A filled example for a typical REST API is in the template's\n   comments.\n3. **Run** — host baseline first, then the cap:\n   ```bash\n   # host (uncapped upper bound):\n   SERVER_ENTRY=src\u002Fmain.ts PORT=3000 DB_PATH=\u002Ftmp\u002Fapp.loadtest.db \\\n     LOADTEST_PROFILE=mixed LOADTEST_SLO_P99_MS=200 \\\n     node --import tsx .codespine\u002Fworkload\u002Floadtest_driver.ts\n\n   # docker (one constrained \"server\" box — realism):\n   docker build -t codespine-workload-runner .codespine\u002Fworkload\n   docker run --rm --cpus 0.5 --memory 512m --memory-swap 512m \\\n     -v \"$PWD:\u002Fwork:ro\" -v \"$PWD\u002F.codespine\u002Fworkload:\u002Fopt\u002Frunner\u002Fwl\" \\\n     -e SERVER_ENTRY=\u002Fwork\u002Fsrc\u002Fmain.ts -e SERVER_CWD=\u002Fopt\u002Frunner \\\n     -e PORT=3000 -e DB_PATH=\u002Ftmp\u002Fapp.loadtest.db \\\n     -e LOADTEST_PROFILE=mixed -e LOADTEST_SLO_P99_MS=200 \\\n     codespine-workload-runner \\\n     node --import tsx \u002Fopt\u002Frunner\u002Fwl\u002Floadtest_driver.ts\n   ```\n4. **Read the ramp**: the driver prints a table (offered rps → p50\u002Fp99) and the\n   knee. **Capacity = the highest sustained rps with p99 under the SLO.** \"Time to\n   add a server\" = steady-state load approaching that number.\n5. **Critical**: use a **file-backed** datastore for any write-inclusive profile\n   (set `DB_PATH`\u002Fequivalent to a file). With an in-memory store, fsync is a no-op\n   and write capacity is fictitious.\n\n## Choosing host vs. docker\n\n- **Host** — uncapped, uses the project's own `node_modules`, no container friction.\n  The easy starting point and the *upper-bound* baseline. A single-threaded server\n  effectively uses ~one core regardless of host core count.\n- **Docker** — an enforced cgroup cap is the only way to get a faithful \"one\n  server\" box (macOS has no native per-process hard cap). Use it for the realism\n  answer. Co-locate the load client in the container against `127.0.0.1` (avoids the\n  host↔VM network-latency floor; the trade-off is the client shares the CPU cap).\n  If the project has runtime dependencies with **native addons**, the container\n  needs a *Linux* build of them (a macOS-arm64 `node_modules` will not run in the\n  Linux VM) — provision them into a volume or install in the image. When that is\n  painful, prefer host mode.\n\nRun both and compare: the gap tells you how much of the slowness is the cap (cgroup\nthrottling + a squeezed page cache) versus the program itself.\n\n## Interpreting and reporting\n\n- Always attach the **realism caveat** to capped numbers (noise; estimate not gate).\n- At low offered rates, **p99 is sample-thin** (few requests) — lengthen the step\n  duration for a stable tail, and say so.\n- A *sharp* knee (throughput plateaus while latency explodes) usually means a\n  **serial bottleneck** (e.g. a synchronous DB on one event loop) — more cores will\n  not help a single-threaded server; \"more servers\" means more processes\u002Freplicas.\n- Tie findings to the graph: `cpu-profile` → name the hot symbols from `hotspots`;\n  for a slow endpoint, read its handler → service → datastore path.\n\n## Limitations \u002F when to escalate\n\nThis MVP assumes the workload is runnable via `node`\u002F`tsx`. Projects needing a\nbuild step, a non-npm package manager, non-HTTP load, custom readiness, or auth\nneed the driver\u002Frun-command adapted by hand — note that to the user rather than\nreporting a bogus number. For cpu-profile, `codespine workload run` (host and\n`--docker`) and `codespine workload scaffold` already wrap this orchestration; the\n`loadtest` kind is run from the scaffolded driver directly for now. The cpu-profile\n`--docker` path auto-provisions a Linux build of the project's dependencies into a\ncached volume (keyed by the lockfile); the loadtest kind under `--docker` is not\nwired yet (use host).\n",{"name":5,"description":7},"main","112dda9bf68462c7fbd2a86d8549d92e359106bf","dotclaude_folder\u002Fskills\u002Fcodespine-workload",0,"https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine","https:\u002F\u002Fjeromeetienne.github.io\u002Fcodespine\u002F","https:\u002F\u002Fgithub.com\u002Fjeromeetienne","http:\u002F\u002Ftwitter.com\u002Fjerome_etienne","I build AI tools that turn ideas into shipped products in days.","2026-06-29 16:35:16",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":46},[36,39,42],{"tool":37,"command":38},"npx","npx skills add jeromeetienne\u002Fcodespine --skill codespine-workload",{"tool":40,"command":41},"bunx","bunx skills add jeromeetienne\u002Fcodespine --skill codespine-workload",{"tool":43,"command":44},"pnpm","pnpm dlx skills add jeromeetienne\u002Fcodespine --skill codespine-workload","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-workload\u002FSKILL.md",true,[48,54,60,66],{"path":49,"resourceType":50,"resourceTypeDesc":51,"sizeBytes":52,"sourceHtmlUrl":53},"reference\u002Ftwo-track.md","REFERENCE","参考文档",2873,"https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-workload\u002Freference\u002Ftwo-track.md",{"path":55,"resourceType":56,"resourceTypeDesc":57,"sizeBytes":58,"sourceHtmlUrl":59},"templates\u002Fcpu_profile_driver.template.ts","SCRIPT","脚本",2098,"https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-workload\u002Ftemplates\u002Fcpu_profile_driver.template.ts",{"path":61,"resourceType":62,"resourceTypeDesc":63,"sizeBytes":64,"sourceHtmlUrl":65},"templates\u002FDockerfile","OTHER","其他",1050,"https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-workload\u002Ftemplates\u002FDockerfile",{"path":67,"resourceType":56,"resourceTypeDesc":57,"sizeBytes":68,"sourceHtmlUrl":69},"templates\u002Floadtest_driver.template.ts",13908,"https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-workload\u002Ftemplates\u002Floadtest_driver.template.ts",[71,78],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":75,"categories":76,"keywords":77,"publishedAt":19},"01M75JP81ZN42PKAKFF883NNH4","codespine-query","Query a TypeScript codebase as a knowledge graph to answer impact, caller, dependency, and dead-code questions. Use this instead of grep\u002Fglob for \"who calls X\", \"what breaks if I change X\" (blast radius), \"is X dead code\", \"what references this type\", and \"what is X connected to\". Requires the codespine CLI and a built graph database.","https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002Fdotclaude_folder\u002Fskills\u002Fcodespine-query\u002FSKILL.md",[],[],{"skillNo":79,"name":73,"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":80,"categories":81,"keywords":82,"publishedAt":19},"01M75JP6Q9NYZKST6E9YM7548V","https:\u002F\u002Fgithub.com\u002Fjeromeetienne\u002Fcodespine\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Fcodespine-query\u002FSKILL.md",[],[]]