[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M753G3KHSDBFFZ1037Z147Z2":3,"skill-same:01M753G3KHSDBFFZ1037Z147Z2":74},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"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":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":32,"install":33,"resources":45},"01M753G3KHSDBFFZ1037Z147Z2","hosted-compute-and-integrations",null,"Routes hosted Jobs and Sandboxes, Space runtime configuration, OAuth and webhooks, framework\u002Fmodel integration, repository cards, tensor serialization, and TensorBoard logging workflows.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: hosted-compute-and-integrations\ndescription: \"Routes hosted Jobs and Sandboxes, Space runtime configuration, OAuth and webhooks, framework\u002Fmodel integration, repository cards, tensor serialization, and TensorBoard logging workflows.\"\ndisable-model-invocation: true\nlicense: Apache-2.0\nmetadata:\n  disco-role: operating\n---\n\n# Hosted compute and integrations\n\nUse this route when a task mentions Hugging Face Jobs, scheduled Jobs, a\nSandbox or SandboxPool, Space hardware\u002Fstorage\u002Fsecrets\u002Fvariables, OAuth,\nWebhooksServer, model-framework integration, model\u002Fdataset\u002FSpace cards, DDUF,\nsharded checkpoints, safetensors, or TensorBoard traces.\n\n## Choose the route\n\n- Read [jobs-and-sandboxes.md](references\u002Fjobs-and-sandboxes.md) for Jobs,\n  scheduled Jobs, hardware selection, volumes, logs, and sandbox lifecycle.\n- Read [spaces-and-server-integrations.md](references\u002Fspaces-and-server-integrations.md)\n  for Space runtime changes, secrets\u002Fvariables, OAuth, webhook servers, and\n  Hub-managed webhook registration.\n- Read [model-integration-and-cards.md](references\u002Fmodel-integration-and-cards.md)\n  for `ModelHubMixin`, `PyTorchModelHubMixin`, local save\u002Fload, cards,\n  model-index metadata, and the local-versus-Hub TensorBoard writer boundary.\n- Read [serialization.md](references\u002Fserialization.md) for DDUF, torch\n  safetensors\u002Fpickle formats, sharding, index-file safety, and local round trips.\n- Read [troubleshooting.md](references\u002Ftroubleshooting.md) for optional\n  dependencies, invalid resources\u002Fconfiguration, auth\u002Fnetwork errors, and\n  recovery decisions. It is also the route for distinguishing client wait\n  timeouts from server Job\u002FSpace failures.\n\nRun the safe local integration fixture with\n[`scripts\u002Flocal_integration_smoke.py`](scripts\u002Flocal_integration_smoke.py)\nbefore attempting any hosted operation. It creates a temporary model\u002Fcard,\nsharded checkpoint, DDUF archive, and mocked configuration recovery; it does\nnot contact the Hub or change a remote resource.\n\n## Operating contract\n\n1. **Classify side effects first.** `run_job`, scheduled-Job creation or\n   triggering, `Sandbox.create`, `SandboxPool` warm-up, Space creation\u002Fupload,\n   Space hardware\u002Fstorage\u002Fsecret\u002Fvariable changes, webhook registration, card\n   publishing, `push_to_hub`, and `HFSummaryWriter` are credentialed or\n   side-effecting. Treat GPU\u002Fpaid flavors, storage, and long-running processes\n   as potentially billable. Use mocks or local fixtures until the user has\n   explicitly authorized a real target.\n2. **Pin the package contract.** Confirm the installed version and inspect the\n   public signature before composing a command. Representative public APIs are\n   listed in the linked references; do not infer new parameters from an old\n   example. Experimental Sandbox, UV-job, OAuth, WebhooksServer, DDUF, and\n   TensorBoard APIs can change, so pin a compatible `huggingface_hub` version.\n3. **Separate configuration from execution.** Validate image, command, cron,\n   volume URI, model\u002Fcard metadata, checkpoint index, webhook route, and Space\n   settings locally. Keep ordinary environment values in `env`\u002Fvariables and\n   sensitive values in encrypted Job\u002FSpace secrets where supported. Never print\n   secret values or forward an HF token unless it is an explicit requirement.\n4. **Choose resources deliberately.** Query available Jobs\u002FSpace hardware\n   rather than hard-coding an assumption. A dedicated Sandbox supports GPU and\n   stronger VM isolation; a SandboxPool is CPU-only, shares a host, accepts no\n   per-sandbox volume, and does not provide the dedicated Job secret channel.\n   Stop or pause paid resources when the work ends.\n5. **Observe state, do not assume success.** Jobs move through `SCHEDULING` and\n   `RUNNING` to `COMPLETED`, `CANCELED`, `ERROR`, or `DELETED`; `wait_for_job`\n   returns failed final info rather than raising for a failed Job. Spaces may\n   remain in build\u002Fstart stages and need `wait_for_space`; inspect the returned\n   stage and logs before retrying. A scheduled Job's immediate trigger does not\n   alter its schedule.\n6. **Prefer safe serialization.** Save torch weights as safetensors by default,\n   validate the generated index and weight map, load with an explicit device and\n   strictness policy, and treat pickle loading as unsafe for untrusted files.\n   DDUF is a constrained diffusion archive, not a general-purpose ZIP.\n7. **Preserve recovery evidence.** Record the input resource, requested versus\n   actual state, credential scope, error\u002Fstatus message, and the corrective\n   change. After a restart or webhook\u002FOAuth change, re-read runtime\u002Froute state\n   and perform a local or mocked request before repeating a remote mutation.\n\n## Verification route\n\n- Start with import\u002Fsignature checks and the bundled local fixture.\n- Assert local model config, card metadata, tensor keys\u002Fshapes, generated shard\n  index paths, DDUF entry names, and round-trip values.\n- For lifecycle logic, replace `HfApi` methods with autospecced mocks and feed\n  staged `JobInfo`\u002F`SpaceRuntime` results. Assert invalid configuration causes\n  zero client calls, then assert the corrected plan calls only the mocks.\n- Keep socket access blocked while the fixture runs. A green local result proves\n  payload\u002Fconfig handling, not credentials, service capacity, billing, or a\n  production callback. Escalate to live verification only with authorization.\n\n## Minimal environment check\n\nFor a normal install, check only what the requested route needs:\n\n```bash\npython -c \"import huggingface_hub; print(huggingface_hub.__version__)\"\n# Add torch and safetensors for serialization; add gradio for WebhooksServer;\n# add the oauth extra for OAuth; add tensorboard or tensorboardX for logging.\n```\n\nDo not run the following during safe verification: hosted Job\u002FSandbox creation,\n`SandboxPool` construction with warm-up, Space or endpoint resource changes,\nsecret writes\u002Fdeletes, webhook create\u002Fupdate\u002Fdelete, card\u002Fmixin uploads,\n`push_to_hub`, live OAuth callbacks, `HFSummaryWriter` against a real repo, or\nremote validation\u002Fdownloads. Mock the client boundary and use local fixtures;\nsee [troubleshooting.md](references\u002Ftroubleshooting.md) when a live check is\nexplicitly authorized.\n","# Hosted compute and integrations\n\nUse this route when a task mentions Hugging Face Jobs, scheduled Jobs, a\nSandbox or SandboxPool, Space hardware\u002Fstorage\u002Fsecrets\u002Fvariables, OAuth,\nWebhooksServer, model-framework integration, model\u002Fdataset\u002FSpace cards, DDUF,\nsharded checkpoints, safetensors, or TensorBoard traces.\n\n## Choose the route\n\n- Read [jobs-and-sandboxes.md](references\u002Fjobs-and-sandboxes.md) for Jobs,\n  scheduled Jobs, hardware selection, volumes, logs, and sandbox lifecycle.\n- Read [spaces-and-server-integrations.md](references\u002Fspaces-and-server-integrations.md)\n  for Space runtime changes, secrets\u002Fvariables, OAuth, webhook servers, and\n  Hub-managed webhook registration.\n- Read [model-integration-and-cards.md](references\u002Fmodel-integration-and-cards.md)\n  for `ModelHubMixin`, `PyTorchModelHubMixin`, local save\u002Fload, cards,\n  model-index metadata, and the local-versus-Hub TensorBoard writer boundary.\n- Read [serialization.md](references\u002Fserialization.md) for DDUF, torch\n  safetensors\u002Fpickle formats, sharding, index-file safety, and local round trips.\n- Read [troubleshooting.md](references\u002Ftroubleshooting.md) for optional\n  dependencies, invalid resources\u002Fconfiguration, auth\u002Fnetwork errors, and\n  recovery decisions. It is also the route for distinguishing client wait\n  timeouts from server Job\u002FSpace failures.\n\nRun the safe local integration fixture with\n[`scripts\u002Flocal_integration_smoke.py`](scripts\u002Flocal_integration_smoke.py)\nbefore attempting any hosted operation. It creates a temporary model\u002Fcard,\nsharded checkpoint, DDUF archive, and mocked configuration recovery; it does\nnot contact the Hub or change a remote resource.\n\n## Operating contract\n\n1. **Classify side effects first.** `run_job`, scheduled-Job creation or\n   triggering, `Sandbox.create`, `SandboxPool` warm-up, Space creation\u002Fupload,\n   Space hardware\u002Fstorage\u002Fsecret\u002Fvariable changes, webhook registration, card\n   publishing, `push_to_hub`, and `HFSummaryWriter` are credentialed or\n   side-effecting. Treat GPU\u002Fpaid flavors, storage, and long-running processes\n   as potentially billable. Use mocks or local fixtures until the user has\n   explicitly authorized a real target.\n2. **Pin the package contract.** Confirm the installed version and inspect the\n   public signature before composing a command. Representative public APIs are\n   listed in the linked references; do not infer new parameters from an old\n   example. Experimental Sandbox, UV-job, OAuth, WebhooksServer, DDUF, and\n   TensorBoard APIs can change, so pin a compatible `huggingface_hub` version.\n3. **Separate configuration from execution.** Validate image, command, cron,\n   volume URI, model\u002Fcard metadata, checkpoint index, webhook route, and Space\n   settings locally. Keep ordinary environment values in `env`\u002Fvariables and\n   sensitive values in encrypted Job\u002FSpace secrets where supported. Never print\n   secret values or forward an HF token unless it is an explicit requirement.\n4. **Choose resources deliberately.** Query available Jobs\u002FSpace hardware\n   rather than hard-coding an assumption. A dedicated Sandbox supports GPU and\n   stronger VM isolation; a SandboxPool is CPU-only, shares a host, accepts no\n   per-sandbox volume, and does not provide the dedicated Job secret channel.\n   Stop or pause paid resources when the work ends.\n5. **Observe state, do not assume success.** Jobs move through `SCHEDULING` and\n   `RUNNING` to `COMPLETED`, `CANCELED`, `ERROR`, or `DELETED`; `wait_for_job`\n   returns failed final info rather than raising for a failed Job. Spaces may\n   remain in build\u002Fstart stages and need `wait_for_space`; inspect the returned\n   stage and logs before retrying. A scheduled Job's immediate trigger does not\n   alter its schedule.\n6. **Prefer safe serialization.** Save torch weights as safetensors by default,\n   validate the generated index and weight map, load with an explicit device and\n   strictness policy, and treat pickle loading as unsafe for untrusted files.\n   DDUF is a constrained diffusion archive, not a general-purpose ZIP.\n7. **Preserve recovery evidence.** Record the input resource, requested versus\n   actual state, credential scope, error\u002Fstatus message, and the corrective\n   change. After a restart or webhook\u002FOAuth change, re-read runtime\u002Froute state\n   and perform a local or mocked request before repeating a remote mutation.\n\n## Verification route\n\n- Start with import\u002Fsignature checks and the bundled local fixture.\n- Assert local model config, card metadata, tensor keys\u002Fshapes, generated shard\n  index paths, DDUF entry names, and round-trip values.\n- For lifecycle logic, replace `HfApi` methods with autospecced mocks and feed\n  staged `JobInfo`\u002F`SpaceRuntime` results. Assert invalid configuration causes\n  zero client calls, then assert the corrected plan calls only the mocks.\n- Keep socket access blocked while the fixture runs. A green local result proves\n  payload\u002Fconfig handling, not credentials, service capacity, billing, or a\n  production callback. Escalate to live verification only with authorization.\n\n## Minimal environment check\n\nFor a normal install, check only what the requested route needs:\n\n```bash\npython -c \"import huggingface_hub; print(huggingface_hub.__version__)\"\n# Add torch and safetensors for serialization; add gradio for WebhooksServer;\n# add the oauth extra for OAuth; add tensorboard or tensorboardX for logging.\n```\n\nDo not run the following during safe verification: hosted Job\u002FSandbox creation,\n`SandboxPool` construction with warm-up, Space or endpoint resource changes,\nsecret writes\u002Fdeletes, webhook create\u002Fupdate\u002Fdelete, card\u002Fmixin uploads,\n`push_to_hub`, live OAuth callbacks, `HFSummaryWriter` against a real repo, or\nremote validation\u002Fdownloads. Mock the client boundary and use local fixtures;\nsee [troubleshooting.md](references\u002Ftroubleshooting.md) when a live check is\nexplicitly authorized.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","examples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":25},[35,38,41],{"tool":36,"command":37},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill hosted-compute-and-integrations",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill hosted-compute-and-integrations",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill hosted-compute-and-integrations","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002FSKILL.md",[46,52,56,60,64,68],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fjobs-and-sandboxes.md","REFERENCE","参考文档",13865,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002Freferences\u002Fjobs-and-sandboxes.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fmodel-integration-and-cards.md",10840,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002Freferences\u002Fmodel-integration-and-cards.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Fserialization.md",8740,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002Freferences\u002Fserialization.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Fspaces-and-server-integrations.md",11279,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002Freferences\u002Fspaces-and-server-integrations.md",{"path":65,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":66,"sourceHtmlUrl":67},"references\u002Ftroubleshooting.md",21929,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002Freferences\u002Ftroubleshooting.md",{"path":69,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":72,"sourceHtmlUrl":73},"scripts\u002Flocal_integration_smoke.py","SCRIPT","脚本",12284,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fexamples\u002Fcreator\u002Frepo-to-skills\u002Fartifacts\u002Fhuggingface_hub\u002Fhuggingface-hub\u002Fsub-skills\u002Fhosted-compute-and-integrations\u002Fscripts\u002Flocal_integration_smoke.py",[75,83,90,97,104,111],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":82},"01M799NYSH1R3RVAHSM2D88Y84","passnet-triton-opt","Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block\u002Fgrid\u002Fwarp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-triton-opt\u002FSKILL.md",[],[],"2026-11-15 14:19:17",{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"publishedAt":82},"01M799NS5APEC0N1FZ7XZVW6BN","passnet-solve","END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-solve\u002FSKILL.md",[],[],{"skillNo":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"publishedAt":82},"01M799NNY9BKSM21V5R7QVTBJX","passnet-skill","PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under .\u002Fpass_dir\u002F, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-skill\u002FSKILL.md",[],[],{"skillNo":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":101,"categories":102,"keywords":103,"publishedAt":82},"01M799NHH6NP7Z1RK77KGXAN6S","passnet-pattern-fusion","Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()\u002Freplacement_args()\u002Freplacement_func() files, when a pass \"failed to match\", or when deciding how to fuse multiple kernels\u002Fops.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-pattern-fusion\u002FSKILL.md",[],[],{"skillNo":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":110,"publishedAt":82},"01M799ND2J4WNC2PGAZ0A06NDF","passnet-orchestrate","Round\u002Fbatch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep\u002Frevert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-orchestrate\u002FSKILL.md",[],[],{"skillNo":112,"name":113,"nameZh":6,"description":114,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":115,"categories":116,"keywords":117,"publishedAt":82},"01M799N8SBTMY02ZDG2AZF5ZWE","passnet-feedback","Fast, deterministic feedback for PassNet work: pre-flight pattern\u002Fmatch verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-feedback\u002FSKILL.md",[],[]]