[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-azure-ai-infra":3,"mdc--sngrgg-key":35,"related-repo-azure-ai-infra":1055,"related-org-azure-ai-infra":1065},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":12,"stars":24,"repoUrl":25,"updatedAt":26,"license":27,"forks":28,"topics":29,"repo":30,"sourceUrl":33,"mdContent":34},"ai-infra","manage Azure HPC and AI infrastructure","Azure HPC and AI infrastructure operations on CycleCloud Workspace for Slurm and AKS clusters with NVIDIA GB300\u002FH100 GPUs. Use for: NCCL bandwidth tests, GPU validation, InfiniBand checks, thermal stress, rack topology (MNNVL), Azure Guest Health Reporting (aka GHR and used for reporting a node failure to Azure), draining and replacing bad nodes, outlier detection, and SKU performance baselines. Covers both Slurm and AKS execution paths.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},"azure","Azure (Microsoft)","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fazure.png","Azure",[13,17,18,21],{"name":14,"slug":15,"type":16},"Operations","operations","tag",{"name":11,"slug":8,"type":16},{"name":19,"slug":20,"type":16},"AI Infrastructure","ai-infrastructure",{"name":22,"slug":23,"type":16},"Kubernetes","kubernetes",31,"https:\u002F\u002Fgithub.com\u002FAzure\u002Fai-infrastructure-on-azure","2026-07-12T08:18:16.298894",null,15,[],{"repoUrl":25,"stars":24,"forks":28,"topics":31,"description":32},[],"This repositories contains examples and best practices for AI workloads on Azure","https:\u002F\u002Fgithub.com\u002FAzure\u002Fai-infrastructure-on-azure\u002Ftree\u002FHEAD\u002Fskills\u002Fai-infra","---\nname: ai-infra\ndescription: \"Azure HPC and AI infrastructure operations on CycleCloud Workspace for Slurm and AKS clusters with NVIDIA GB300\u002FH100 GPUs. Use for: NCCL bandwidth tests, GPU validation, InfiniBand checks, thermal stress, rack topology (MNNVL), Azure Guest Health Reporting (aka GHR and used for reporting a node failure to Azure), draining and replacing bad nodes, outlier detection, and SKU performance baselines. Covers both Slurm and AKS execution paths.\"\n---\n\n# AI Infrastructure on Azure\n\nOperational knowledge for Azure HPC GPU clusters running on CycleCloud Workspace for Slurm or AKS, with NVIDIA GB300 (NDv6) and H100 (NDv5) GPUs.\n\nThis single skill is the entry point for all cluster validation, diagnosis, and remediation questions. It contains a routing index that points to detailed reference files in `references\u002F`. **Read the relevant reference files before giving commands** — do not invent thresholds or procedures.\n\n## How to Use This Skill\n\n1. **Classify the request** using the Intent Map below.\n2. **Read the listed reference files** before answering.\n3. **Use only commands and thresholds that appear in the reference files.** If the user's question can't be answered from the references, say so and ask for clarification.\n4. **Ask for missing inputs** explicitly: SKU (`Standard_ND128isr_GB300_v6` or `Standard_ND96isr_H100_v5`), nodelist, cluster name, orchestrator (Slurm or AKS), failing job context.\n5. **Respond using the Response Contract** at the bottom.\n\nIf the user invokes this skill with `\u002Fai-infra \u003Crequest>`, treat that as an explicit invocation and proceed with the same workflow.\n\n## Reference File Index\n\nAll reference files live in `skills\u002Fai-infra\u002Freferences\u002F`.\n\n### Concepts and baselines (orchestrator-agnostic)\n\n| File                   | Covers                                                                                                                                                                                  |\n| ---------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| `sku_baselines.md`     | Expected NCCL busbw, GPU GFlops, thermal limits, IB ports, rack sizes for GB300 and H100. Warn and GHR thresholds.                                                                      |\n| `rack_topology.md`     | MNNVL domain discovery via `nvidia-smi`, ClusterUUID, expected rack sizes per SKU, FabricManager troubleshooting.                                                                       |\n| `ib_validation.md`     | InfiniBand port state (`operstate`, `ibstat`), partition keys, error counters, link flap detection, soft fixes.                                                                         |\n| `nccl_diagnosis.md`    | Bisection algorithm, intra-rack vs inter-rack scoping, GPU vs network root cause, bandwidth pattern interpretation.                                                                     |\n| `outlier_detection.md` | Statistical methods (absolute threshold, z-score, MAD) for fleet-wide GPU and NCCL analysis.                                                                                            |\n| `ghr.md`               | **Azure Guest Health Reporting (GHR)** — used for reporting a node failure to Azure. Covers full impact category reference, IMDS\u002FKVP data collection, REST API format, insight polling. |\n\n### Test execution (currently Slurm; AKS counterparts to be added)\n\n| File                | Covers                                                                                                                                                                       |\n| ------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |\n| `nccl_test.md`      | Run `all_reduce_perf` via the Slurm launcher. Per-SKU env vars (MNNVL\u002FSHARP\u002FGDR), output columns, quick vs full sweep. AKS path: see `infrastructure_validations\u002Faks\u002FNCCL\u002F`. |\n| `gpu_validation.md` | Run ubergemm GEMM benchmark via Slurm. Parse CSV output, identify underperforming GPUs.                                                                                      |\n| `thermal_test.md`   | Run dcgmproftester thermal stress via Slurm. Interpret pass\u002Ffail, throttle reasons, DCGMI levels. AKS path: see `infrastructure_validations\u002Faks\u002F`.                           |\n\n### Remediation\n\n| File                        | Covers                                                                                                 |\n| --------------------------- | ------------------------------------------------------------------------------------------------------ |\n| `node_drain_and_replace.md` | Slurm `drain`\u002F`undrain`\u002Freboot, decision tree for drain vs reboot vs GHR, post-replacement validation. |\n\n## Intent Map\n\n### New cluster bring-up \u002F full validation\n\nRead: `sku_baselines.md`, `rack_topology.md`, `nccl_test.md`, `gpu_validation.md`, `thermal_test.md`\n\n### Slow training or low multi-node throughput\n\nRead: `nccl_diagnosis.md`, `sku_baselines.md`, `ib_validation.md`, `rack_topology.md` (when topology correlation is needed)\n\n### NCCL failures or low all-reduce bandwidth\n\nRead: `nccl_test.md`, `nccl_diagnosis.md`, `ib_validation.md`, `outlier_detection.md` (fleet-wide analysis)\n\n### GPU underperformance on one or more nodes\n\nRead: `gpu_validation.md`, `outlier_detection.md`, `sku_baselines.md`\n\n### Thermal throttling or suspected cooling issues\n\nRead: `thermal_test.md`, `sku_baselines.md`, `gpu_validation.md` (if thermal impact on GEMM is suspected)\n\n### InfiniBand link \u002F pkey \u002F errors investigation\n\nRead: `ib_validation.md`, `nccl_diagnosis.md`, `rack_topology.md` (when rack-locality matters)\n\n### Identify degraded nodes across fleet\n\nRead: `outlier_detection.md`, `sku_baselines.md`, plus `gpu_validation.md` and\u002For `nccl_test.md` depending on the metric source\n\n### Node remediation \u002F drain \u002F replace \u002F file GHR (Guest Health Report)\n\nTriggers: \"file a GHR\", \"guest health report\", \"report a bad node to Azure\", \"report a node failure to Azure\", \"open a hardware issue with Azure\", \"health impact category\".\n\nRead: `node_drain_and_replace.md`, `ghr.md`\n\n## Cross-Cutting Rules\n\nThese apply to every operational answer:\n\n1. **Always collect node metadata before rebooting or draining.** PhysicalHostName (Hyper-V KVP pool 3) and Resource ID (IMDS) are required for any GHR. If the node goes down, you lose access to them. See `ghr.md` for exact commands.\n2. **Use exact thresholds from `sku_baselines.md`.** Do not invent warn\u002FGHR cutoffs.\n3. **Never suppress test failures** — non-zero `#wrong` in NCCL output, dcgmproftester non-zero exit, XID errors in dmesg are all hard signals.\n4. **For NCCL bandwidth issues, scope before bisecting.** Determine intra-rack vs inter-rack from per-rack tests before isolating individual nodes.\n5. **GHR impact category must match the symptom.** See the category reference table in `ghr.md`. Use `HpcGenericFailure` only when nothing else fits.\n6. **Test scripts live in `infrastructure_validations\u002F`** in this repo. The references describe how to run them and how to interpret output; the runnable artifacts are the scripts themselves.\n\n## Test Script Paths\n\n- `infrastructure_validations\u002Fslurm\u002FNCCL\u002F` — NCCL `all_reduce_perf` launcher with per-SKU configs\n- `infrastructure_validations\u002Fslurm\u002Fgpu_test\u002F` — GPU GEMM benchmark (ubergemm)\n- `infrastructure_validations\u002Fslurm\u002Fthermal_test\u002F` — Thermal stress test (dcgmproftester)\n- `infrastructure_validations\u002Fslurm\u002FNHC\u002F` — Node Health Check\n- `infrastructure_validations\u002Faks\u002FNCCL\u002F` — NCCL on AKS (via MPI Operator)\n- `infrastructure_validations\u002Faks\u002FNHC\u002F` — Node Health Check on AKS\n- `infrastructure_validations\u002Faks\u002Ffio\u002F` — Storage I\u002FO test on AKS\n\n## Response Contract\n\nFor operational responses, follow this structure:\n\n1. **Selected references** — list the reference files you used.\n2. **Run plan** — concise ordered steps.\n3. **Exact commands** — copied from the reference files.\n4. **Pass\u002Ffail thresholds** — from `sku_baselines.md` or the relevant test reference.\n5. **Action decision** — continue, isolate, drain, reboot, or file GHR with the specific impact category.\n",{"data":36,"body":37},{"name":4,"description":6},{"type":38,"children":39},"root",[40,49,55,77,84,155,168,174,187,194,353,359,452,458,511,517,523,557,563,592,598,627,633,654,660,683,689,712,718,749,755,760,775,781,786,891,897,984,990,995],{"type":41,"tag":42,"props":43,"children":45},"element","h1",{"id":44},"ai-infrastructure-on-azure",[46],{"type":47,"value":48},"text","AI Infrastructure on Azure",{"type":41,"tag":50,"props":51,"children":52},"p",{},[53],{"type":47,"value":54},"Operational knowledge for Azure HPC GPU clusters running on CycleCloud Workspace for Slurm or AKS, with NVIDIA GB300 (NDv6) and H100 (NDv5) GPUs.",{"type":41,"tag":50,"props":56,"children":57},{},[58,60,67,69,75],{"type":47,"value":59},"This single skill is the entry point for all cluster validation, diagnosis, and remediation questions. It contains a routing index that points to detailed reference files in ",{"type":41,"tag":61,"props":62,"children":64},"code",{"className":63},[],[65],{"type":47,"value":66},"references\u002F",{"type":47,"value":68},". ",{"type":41,"tag":70,"props":71,"children":72},"strong",{},[73],{"type":47,"value":74},"Read the relevant reference files before giving commands",{"type":47,"value":76}," — do not invent thresholds or procedures.",{"type":41,"tag":78,"props":79,"children":81},"h2",{"id":80},"how-to-use-this-skill",[82],{"type":47,"value":83},"How to Use This Skill",{"type":41,"tag":85,"props":86,"children":87},"ol",{},[88,99,109,119,145],{"type":41,"tag":89,"props":90,"children":91},"li",{},[92,97],{"type":41,"tag":70,"props":93,"children":94},{},[95],{"type":47,"value":96},"Classify the request",{"type":47,"value":98}," using the Intent Map below.",{"type":41,"tag":89,"props":100,"children":101},{},[102,107],{"type":41,"tag":70,"props":103,"children":104},{},[105],{"type":47,"value":106},"Read the listed reference files",{"type":47,"value":108}," before answering.",{"type":41,"tag":89,"props":110,"children":111},{},[112,117],{"type":41,"tag":70,"props":113,"children":114},{},[115],{"type":47,"value":116},"Use only commands and thresholds that appear in the reference files.",{"type":47,"value":118}," If the user's question can't be answered from the references, say so and ask for clarification.",{"type":41,"tag":89,"props":120,"children":121},{},[122,127,129,135,137,143],{"type":41,"tag":70,"props":123,"children":124},{},[125],{"type":47,"value":126},"Ask for missing inputs",{"type":47,"value":128}," explicitly: SKU (",{"type":41,"tag":61,"props":130,"children":132},{"className":131},[],[133],{"type":47,"value":134},"Standard_ND128isr_GB300_v6",{"type":47,"value":136}," or ",{"type":41,"tag":61,"props":138,"children":140},{"className":139},[],[141],{"type":47,"value":142},"Standard_ND96isr_H100_v5",{"type":47,"value":144},"), nodelist, cluster name, orchestrator (Slurm or AKS), failing job context.",{"type":41,"tag":89,"props":146,"children":147},{},[148,153],{"type":41,"tag":70,"props":149,"children":150},{},[151],{"type":47,"value":152},"Respond using the Response Contract",{"type":47,"value":154}," at the bottom.",{"type":41,"tag":50,"props":156,"children":157},{},[158,160,166],{"type":47,"value":159},"If the user invokes this skill with ",{"type":41,"tag":61,"props":161,"children":163},{"className":162},[],[164],{"type":47,"value":165},"\u002Fai-infra \u003Crequest>",{"type":47,"value":167},", treat that as an explicit invocation and proceed with the same workflow.",{"type":41,"tag":78,"props":169,"children":171},{"id":170},"reference-file-index",[172],{"type":47,"value":173},"Reference File Index",{"type":41,"tag":50,"props":175,"children":176},{},[177,179,185],{"type":47,"value":178},"All reference files live in ",{"type":41,"tag":61,"props":180,"children":182},{"className":181},[],[183],{"type":47,"value":184},"skills\u002Fai-infra\u002Freferences\u002F",{"type":47,"value":186},".",{"type":41,"tag":188,"props":189,"children":191},"h3",{"id":190},"concepts-and-baselines-orchestrator-agnostic",[192],{"type":47,"value":193},"Concepts and baselines (orchestrator-agnostic)",{"type":41,"tag":195,"props":196,"children":197},"table",{},[198,217],{"type":41,"tag":199,"props":200,"children":201},"thead",{},[202],{"type":41,"tag":203,"props":204,"children":205},"tr",{},[206,212],{"type":41,"tag":207,"props":208,"children":209},"th",{},[210],{"type":47,"value":211},"File",{"type":41,"tag":207,"props":213,"children":214},{},[215],{"type":47,"value":216},"Covers",{"type":41,"tag":218,"props":219,"children":220},"tbody",{},[221,239,264,297,314,331],{"type":41,"tag":203,"props":222,"children":223},{},[224,234],{"type":41,"tag":225,"props":226,"children":227},"td",{},[228],{"type":41,"tag":61,"props":229,"children":231},{"className":230},[],[232],{"type":47,"value":233},"sku_baselines.md",{"type":41,"tag":225,"props":235,"children":236},{},[237],{"type":47,"value":238},"Expected NCCL busbw, GPU GFlops, thermal limits, IB ports, rack sizes for GB300 and H100. Warn and GHR thresholds.",{"type":41,"tag":203,"props":240,"children":241},{},[242,251],{"type":41,"tag":225,"props":243,"children":244},{},[245],{"type":41,"tag":61,"props":246,"children":248},{"className":247},[],[249],{"type":47,"value":250},"rack_topology.md",{"type":41,"tag":225,"props":252,"children":253},{},[254,256,262],{"type":47,"value":255},"MNNVL domain discovery via ",{"type":41,"tag":61,"props":257,"children":259},{"className":258},[],[260],{"type":47,"value":261},"nvidia-smi",{"type":47,"value":263},", ClusterUUID, expected rack sizes per SKU, FabricManager troubleshooting.",{"type":41,"tag":203,"props":265,"children":266},{},[267,276],{"type":41,"tag":225,"props":268,"children":269},{},[270],{"type":41,"tag":61,"props":271,"children":273},{"className":272},[],[274],{"type":47,"value":275},"ib_validation.md",{"type":41,"tag":225,"props":277,"children":278},{},[279,281,287,289,295],{"type":47,"value":280},"InfiniBand port state (",{"type":41,"tag":61,"props":282,"children":284},{"className":283},[],[285],{"type":47,"value":286},"operstate",{"type":47,"value":288},", ",{"type":41,"tag":61,"props":290,"children":292},{"className":291},[],[293],{"type":47,"value":294},"ibstat",{"type":47,"value":296},"), partition keys, error counters, link flap detection, soft fixes.",{"type":41,"tag":203,"props":298,"children":299},{},[300,309],{"type":41,"tag":225,"props":301,"children":302},{},[303],{"type":41,"tag":61,"props":304,"children":306},{"className":305},[],[307],{"type":47,"value":308},"nccl_diagnosis.md",{"type":41,"tag":225,"props":310,"children":311},{},[312],{"type":47,"value":313},"Bisection algorithm, intra-rack vs inter-rack scoping, GPU vs network root cause, bandwidth pattern interpretation.",{"type":41,"tag":203,"props":315,"children":316},{},[317,326],{"type":41,"tag":225,"props":318,"children":319},{},[320],{"type":41,"tag":61,"props":321,"children":323},{"className":322},[],[324],{"type":47,"value":325},"outlier_detection.md",{"type":41,"tag":225,"props":327,"children":328},{},[329],{"type":47,"value":330},"Statistical methods (absolute threshold, z-score, MAD) for fleet-wide GPU and NCCL analysis.",{"type":41,"tag":203,"props":332,"children":333},{},[334,343],{"type":41,"tag":225,"props":335,"children":336},{},[337],{"type":41,"tag":61,"props":338,"children":340},{"className":339},[],[341],{"type":47,"value":342},"ghr.md",{"type":41,"tag":225,"props":344,"children":345},{},[346,351],{"type":41,"tag":70,"props":347,"children":348},{},[349],{"type":47,"value":350},"Azure Guest Health Reporting (GHR)",{"type":47,"value":352}," — used for reporting a node failure to Azure. Covers full impact category reference, IMDS\u002FKVP data collection, REST API format, insight polling.",{"type":41,"tag":188,"props":354,"children":356},{"id":355},"test-execution-currently-slurm-aks-counterparts-to-be-added",[357],{"type":47,"value":358},"Test execution (currently Slurm; AKS counterparts to be added)",{"type":41,"tag":195,"props":360,"children":361},{},[362,376],{"type":41,"tag":199,"props":363,"children":364},{},[365],{"type":41,"tag":203,"props":366,"children":367},{},[368,372],{"type":41,"tag":207,"props":369,"children":370},{},[371],{"type":47,"value":211},{"type":41,"tag":207,"props":373,"children":374},{},[375],{"type":47,"value":216},{"type":41,"tag":218,"props":377,"children":378},{},[379,411,428],{"type":41,"tag":203,"props":380,"children":381},{},[382,391],{"type":41,"tag":225,"props":383,"children":384},{},[385],{"type":41,"tag":61,"props":386,"children":388},{"className":387},[],[389],{"type":47,"value":390},"nccl_test.md",{"type":41,"tag":225,"props":392,"children":393},{},[394,396,402,404,410],{"type":47,"value":395},"Run ",{"type":41,"tag":61,"props":397,"children":399},{"className":398},[],[400],{"type":47,"value":401},"all_reduce_perf",{"type":47,"value":403}," via the Slurm launcher. Per-SKU env vars (MNNVL\u002FSHARP\u002FGDR), output columns, quick vs full sweep. AKS path: see ",{"type":41,"tag":61,"props":405,"children":407},{"className":406},[],[408],{"type":47,"value":409},"infrastructure_validations\u002Faks\u002FNCCL\u002F",{"type":47,"value":186},{"type":41,"tag":203,"props":412,"children":413},{},[414,423],{"type":41,"tag":225,"props":415,"children":416},{},[417],{"type":41,"tag":61,"props":418,"children":420},{"className":419},[],[421],{"type":47,"value":422},"gpu_validation.md",{"type":41,"tag":225,"props":424,"children":425},{},[426],{"type":47,"value":427},"Run ubergemm GEMM benchmark via Slurm. Parse CSV output, identify underperforming GPUs.",{"type":41,"tag":203,"props":429,"children":430},{},[431,440],{"type":41,"tag":225,"props":432,"children":433},{},[434],{"type":41,"tag":61,"props":435,"children":437},{"className":436},[],[438],{"type":47,"value":439},"thermal_test.md",{"type":41,"tag":225,"props":441,"children":442},{},[443,445,451],{"type":47,"value":444},"Run dcgmproftester thermal stress via Slurm. Interpret pass\u002Ffail, throttle reasons, DCGMI levels. AKS path: see ",{"type":41,"tag":61,"props":446,"children":448},{"className":447},[],[449],{"type":47,"value":450},"infrastructure_validations\u002Faks\u002F",{"type":47,"value":186},{"type":41,"tag":188,"props":453,"children":455},{"id":454},"remediation",[456],{"type":47,"value":457},"Remediation",{"type":41,"tag":195,"props":459,"children":460},{},[461,475],{"type":41,"tag":199,"props":462,"children":463},{},[464],{"type":41,"tag":203,"props":465,"children":466},{},[467,471],{"type":41,"tag":207,"props":468,"children":469},{},[470],{"type":47,"value":211},{"type":41,"tag":207,"props":472,"children":473},{},[474],{"type":47,"value":216},{"type":41,"tag":218,"props":476,"children":477},{},[478],{"type":41,"tag":203,"props":479,"children":480},{},[481,490],{"type":41,"tag":225,"props":482,"children":483},{},[484],{"type":41,"tag":61,"props":485,"children":487},{"className":486},[],[488],{"type":47,"value":489},"node_drain_and_replace.md",{"type":41,"tag":225,"props":491,"children":492},{},[493,495,501,503,509],{"type":47,"value":494},"Slurm ",{"type":41,"tag":61,"props":496,"children":498},{"className":497},[],[499],{"type":47,"value":500},"drain",{"type":47,"value":502},"\u002F",{"type":41,"tag":61,"props":504,"children":506},{"className":505},[],[507],{"type":47,"value":508},"undrain",{"type":47,"value":510},"\u002Freboot, decision tree for drain vs reboot vs GHR, post-replacement validation.",{"type":41,"tag":78,"props":512,"children":514},{"id":513},"intent-map",[515],{"type":47,"value":516},"Intent Map",{"type":41,"tag":188,"props":518,"children":520},{"id":519},"new-cluster-bring-up-full-validation",[521],{"type":47,"value":522},"New cluster bring-up \u002F full validation",{"type":41,"tag":50,"props":524,"children":525},{},[526,528,533,534,539,540,545,546,551,552],{"type":47,"value":527},"Read: ",{"type":41,"tag":61,"props":529,"children":531},{"className":530},[],[532],{"type":47,"value":233},{"type":47,"value":288},{"type":41,"tag":61,"props":535,"children":537},{"className":536},[],[538],{"type":47,"value":250},{"type":47,"value":288},{"type":41,"tag":61,"props":541,"children":543},{"className":542},[],[544],{"type":47,"value":390},{"type":47,"value":288},{"type":41,"tag":61,"props":547,"children":549},{"className":548},[],[550],{"type":47,"value":422},{"type":47,"value":288},{"type":41,"tag":61,"props":553,"children":555},{"className":554},[],[556],{"type":47,"value":439},{"type":41,"tag":188,"props":558,"children":560},{"id":559},"slow-training-or-low-multi-node-throughput",[561],{"type":47,"value":562},"Slow training or low multi-node throughput",{"type":41,"tag":50,"props":564,"children":565},{},[566,567,572,573,578,579,584,585,590],{"type":47,"value":527},{"type":41,"tag":61,"props":568,"children":570},{"className":569},[],[571],{"type":47,"value":308},{"type":47,"value":288},{"type":41,"tag":61,"props":574,"children":576},{"className":575},[],[577],{"type":47,"value":233},{"type":47,"value":288},{"type":41,"tag":61,"props":580,"children":582},{"className":581},[],[583],{"type":47,"value":275},{"type":47,"value":288},{"type":41,"tag":61,"props":586,"children":588},{"className":587},[],[589],{"type":47,"value":250},{"type":47,"value":591}," (when topology correlation is needed)",{"type":41,"tag":188,"props":593,"children":595},{"id":594},"nccl-failures-or-low-all-reduce-bandwidth",[596],{"type":47,"value":597},"NCCL failures or low all-reduce bandwidth",{"type":41,"tag":50,"props":599,"children":600},{},[601,602,607,608,613,614,619,620,625],{"type":47,"value":527},{"type":41,"tag":61,"props":603,"children":605},{"className":604},[],[606],{"type":47,"value":390},{"type":47,"value":288},{"type":41,"tag":61,"props":609,"children":611},{"className":610},[],[612],{"type":47,"value":308},{"type":47,"value":288},{"type":41,"tag":61,"props":615,"children":617},{"className":616},[],[618],{"type":47,"value":275},{"type":47,"value":288},{"type":41,"tag":61,"props":621,"children":623},{"className":622},[],[624],{"type":47,"value":325},{"type":47,"value":626}," (fleet-wide analysis)",{"type":41,"tag":188,"props":628,"children":630},{"id":629},"gpu-underperformance-on-one-or-more-nodes",[631],{"type":47,"value":632},"GPU underperformance on one or more nodes",{"type":41,"tag":50,"props":634,"children":635},{},[636,637,642,643,648,649],{"type":47,"value":527},{"type":41,"tag":61,"props":638,"children":640},{"className":639},[],[641],{"type":47,"value":422},{"type":47,"value":288},{"type":41,"tag":61,"props":644,"children":646},{"className":645},[],[647],{"type":47,"value":325},{"type":47,"value":288},{"type":41,"tag":61,"props":650,"children":652},{"className":651},[],[653],{"type":47,"value":233},{"type":41,"tag":188,"props":655,"children":657},{"id":656},"thermal-throttling-or-suspected-cooling-issues",[658],{"type":47,"value":659},"Thermal throttling or suspected cooling issues",{"type":41,"tag":50,"props":661,"children":662},{},[663,664,669,670,675,676,681],{"type":47,"value":527},{"type":41,"tag":61,"props":665,"children":667},{"className":666},[],[668],{"type":47,"value":439},{"type":47,"value":288},{"type":41,"tag":61,"props":671,"children":673},{"className":672},[],[674],{"type":47,"value":233},{"type":47,"value":288},{"type":41,"tag":61,"props":677,"children":679},{"className":678},[],[680],{"type":47,"value":422},{"type":47,"value":682}," (if thermal impact on GEMM is suspected)",{"type":41,"tag":188,"props":684,"children":686},{"id":685},"infiniband-link-pkey-errors-investigation",[687],{"type":47,"value":688},"InfiniBand link \u002F pkey \u002F errors investigation",{"type":41,"tag":50,"props":690,"children":691},{},[692,693,698,699,704,705,710],{"type":47,"value":527},{"type":41,"tag":61,"props":694,"children":696},{"className":695},[],[697],{"type":47,"value":275},{"type":47,"value":288},{"type":41,"tag":61,"props":700,"children":702},{"className":701},[],[703],{"type":47,"value":308},{"type":47,"value":288},{"type":41,"tag":61,"props":706,"children":708},{"className":707},[],[709],{"type":47,"value":250},{"type":47,"value":711}," (when rack-locality matters)",{"type":41,"tag":188,"props":713,"children":715},{"id":714},"identify-degraded-nodes-across-fleet",[716],{"type":47,"value":717},"Identify degraded nodes across fleet",{"type":41,"tag":50,"props":719,"children":720},{},[721,722,727,728,733,735,740,742,747],{"type":47,"value":527},{"type":41,"tag":61,"props":723,"children":725},{"className":724},[],[726],{"type":47,"value":325},{"type":47,"value":288},{"type":41,"tag":61,"props":729,"children":731},{"className":730},[],[732],{"type":47,"value":233},{"type":47,"value":734},", plus ",{"type":41,"tag":61,"props":736,"children":738},{"className":737},[],[739],{"type":47,"value":422},{"type":47,"value":741}," and\u002For ",{"type":41,"tag":61,"props":743,"children":745},{"className":744},[],[746],{"type":47,"value":390},{"type":47,"value":748}," depending on the metric source",{"type":41,"tag":188,"props":750,"children":752},{"id":751},"node-remediation-drain-replace-file-ghr-guest-health-report",[753],{"type":47,"value":754},"Node remediation \u002F drain \u002F replace \u002F file GHR (Guest Health Report)",{"type":41,"tag":50,"props":756,"children":757},{},[758],{"type":47,"value":759},"Triggers: \"file a GHR\", \"guest health report\", \"report a bad node to Azure\", \"report a node failure to Azure\", \"open a hardware issue with Azure\", \"health impact category\".",{"type":41,"tag":50,"props":761,"children":762},{},[763,764,769,770],{"type":47,"value":527},{"type":41,"tag":61,"props":765,"children":767},{"className":766},[],[768],{"type":47,"value":489},{"type":47,"value":288},{"type":41,"tag":61,"props":771,"children":773},{"className":772},[],[774],{"type":47,"value":342},{"type":41,"tag":78,"props":776,"children":778},{"id":777},"cross-cutting-rules",[779],{"type":47,"value":780},"Cross-Cutting Rules",{"type":41,"tag":50,"props":782,"children":783},{},[784],{"type":47,"value":785},"These apply to every operational answer:",{"type":41,"tag":85,"props":787,"children":788},{},[789,806,822,840,850,875],{"type":41,"tag":89,"props":790,"children":791},{},[792,797,799,804],{"type":41,"tag":70,"props":793,"children":794},{},[795],{"type":47,"value":796},"Always collect node metadata before rebooting or draining.",{"type":47,"value":798}," PhysicalHostName (Hyper-V KVP pool 3) and Resource ID (IMDS) are required for any GHR. If the node goes down, you lose access to them. See ",{"type":41,"tag":61,"props":800,"children":802},{"className":801},[],[803],{"type":47,"value":342},{"type":47,"value":805}," for exact commands.",{"type":41,"tag":89,"props":807,"children":808},{},[809,820],{"type":41,"tag":70,"props":810,"children":811},{},[812,814,819],{"type":47,"value":813},"Use exact thresholds from ",{"type":41,"tag":61,"props":815,"children":817},{"className":816},[],[818],{"type":47,"value":233},{"type":47,"value":186},{"type":47,"value":821}," Do not invent warn\u002FGHR cutoffs.",{"type":41,"tag":89,"props":823,"children":824},{},[825,830,832,838],{"type":41,"tag":70,"props":826,"children":827},{},[828],{"type":47,"value":829},"Never suppress test failures",{"type":47,"value":831}," — non-zero ",{"type":41,"tag":61,"props":833,"children":835},{"className":834},[],[836],{"type":47,"value":837},"#wrong",{"type":47,"value":839}," in NCCL output, dcgmproftester non-zero exit, XID errors in dmesg are all hard signals.",{"type":41,"tag":89,"props":841,"children":842},{},[843,848],{"type":41,"tag":70,"props":844,"children":845},{},[846],{"type":47,"value":847},"For NCCL bandwidth issues, scope before bisecting.",{"type":47,"value":849}," Determine intra-rack vs inter-rack from per-rack tests before isolating individual nodes.",{"type":41,"tag":89,"props":851,"children":852},{},[853,858,860,865,867,873],{"type":41,"tag":70,"props":854,"children":855},{},[856],{"type":47,"value":857},"GHR impact category must match the symptom.",{"type":47,"value":859}," See the category reference table in ",{"type":41,"tag":61,"props":861,"children":863},{"className":862},[],[864],{"type":47,"value":342},{"type":47,"value":866},". Use ",{"type":41,"tag":61,"props":868,"children":870},{"className":869},[],[871],{"type":47,"value":872},"HpcGenericFailure",{"type":47,"value":874}," only when nothing else fits.",{"type":41,"tag":89,"props":876,"children":877},{},[878,889],{"type":41,"tag":70,"props":879,"children":880},{},[881,883],{"type":47,"value":882},"Test scripts live in ",{"type":41,"tag":61,"props":884,"children":886},{"className":885},[],[887],{"type":47,"value":888},"infrastructure_validations\u002F",{"type":47,"value":890}," in this repo. The references describe how to run them and how to interpret output; the runnable artifacts are the scripts themselves.",{"type":41,"tag":78,"props":892,"children":894},{"id":893},"test-script-paths",[895],{"type":47,"value":896},"Test Script Paths",{"type":41,"tag":898,"props":899,"children":900},"ul",{},[901,919,930,941,952,962,973],{"type":41,"tag":89,"props":902,"children":903},{},[904,910,912,917],{"type":41,"tag":61,"props":905,"children":907},{"className":906},[],[908],{"type":47,"value":909},"infrastructure_validations\u002Fslurm\u002FNCCL\u002F",{"type":47,"value":911}," — NCCL ",{"type":41,"tag":61,"props":913,"children":915},{"className":914},[],[916],{"type":47,"value":401},{"type":47,"value":918}," launcher with per-SKU configs",{"type":41,"tag":89,"props":920,"children":921},{},[922,928],{"type":41,"tag":61,"props":923,"children":925},{"className":924},[],[926],{"type":47,"value":927},"infrastructure_validations\u002Fslurm\u002Fgpu_test\u002F",{"type":47,"value":929}," — GPU GEMM benchmark (ubergemm)",{"type":41,"tag":89,"props":931,"children":932},{},[933,939],{"type":41,"tag":61,"props":934,"children":936},{"className":935},[],[937],{"type":47,"value":938},"infrastructure_validations\u002Fslurm\u002Fthermal_test\u002F",{"type":47,"value":940}," — Thermal stress test (dcgmproftester)",{"type":41,"tag":89,"props":942,"children":943},{},[944,950],{"type":41,"tag":61,"props":945,"children":947},{"className":946},[],[948],{"type":47,"value":949},"infrastructure_validations\u002Fslurm\u002FNHC\u002F",{"type":47,"value":951}," — Node Health Check",{"type":41,"tag":89,"props":953,"children":954},{},[955,960],{"type":41,"tag":61,"props":956,"children":958},{"className":957},[],[959],{"type":47,"value":409},{"type":47,"value":961}," — NCCL on AKS (via MPI Operator)",{"type":41,"tag":89,"props":963,"children":964},{},[965,971],{"type":41,"tag":61,"props":966,"children":968},{"className":967},[],[969],{"type":47,"value":970},"infrastructure_validations\u002Faks\u002FNHC\u002F",{"type":47,"value":972}," — Node Health Check on AKS",{"type":41,"tag":89,"props":974,"children":975},{},[976,982],{"type":41,"tag":61,"props":977,"children":979},{"className":978},[],[980],{"type":47,"value":981},"infrastructure_validations\u002Faks\u002Ffio\u002F",{"type":47,"value":983}," — Storage I\u002FO test on AKS",{"type":41,"tag":78,"props":985,"children":987},{"id":986},"response-contract",[988],{"type":47,"value":989},"Response Contract",{"type":41,"tag":50,"props":991,"children":992},{},[993],{"type":47,"value":994},"For operational responses, follow this structure:",{"type":41,"tag":85,"props":996,"children":997},{},[998,1008,1018,1028,1045],{"type":41,"tag":89,"props":999,"children":1000},{},[1001,1006],{"type":41,"tag":70,"props":1002,"children":1003},{},[1004],{"type":47,"value":1005},"Selected references",{"type":47,"value":1007}," — list the reference files you used.",{"type":41,"tag":89,"props":1009,"children":1010},{},[1011,1016],{"type":41,"tag":70,"props":1012,"children":1013},{},[1014],{"type":47,"value":1015},"Run plan",{"type":47,"value":1017}," — concise ordered steps.",{"type":41,"tag":89,"props":1019,"children":1020},{},[1021,1026],{"type":41,"tag":70,"props":1022,"children":1023},{},[1024],{"type":47,"value":1025},"Exact commands",{"type":47,"value":1027}," — copied from the reference files.",{"type":41,"tag":89,"props":1029,"children":1030},{},[1031,1036,1038,1043],{"type":41,"tag":70,"props":1032,"children":1033},{},[1034],{"type":47,"value":1035},"Pass\u002Ffail thresholds",{"type":47,"value":1037}," — from ",{"type":41,"tag":61,"props":1039,"children":1041},{"className":1040},[],[1042],{"type":47,"value":233},{"type":47,"value":1044}," or the relevant test reference.",{"type":41,"tag":89,"props":1046,"children":1047},{},[1048,1053],{"type":41,"tag":70,"props":1049,"children":1050},{},[1051],{"type":47,"value":1052},"Action decision",{"type":47,"value":1054}," — continue, isolate, drain, reboot, or file GHR with the specific impact category.",{"items":1056,"total":1064},[1057],{"slug":4,"name":4,"fn":5,"description":6,"org":1058,"tags":1059,"stars":24,"repoUrl":25,"updatedAt":26},{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1060,1061,1062,1063],{"name":19,"slug":20,"type":16},{"name":11,"slug":8,"type":16},{"name":22,"slug":23,"type":16},{"name":14,"slug":15,"type":16},1,{"items":1066,"total":1245},[1067,1086,1105,1124,1139,1154,1167,1182,1193,1207,1220,1233],{"slug":1068,"name":1068,"fn":1069,"description":1070,"org":1071,"tags":1072,"stars":1083,"repoUrl":1084,"updatedAt":1085},"azure-arg-external-evaluation-policy-author","author and test Azure Resource Graph policies","Use when the user wants to author, design, or test an Azure Policy that queries Azure Resource Graph (ARG) at request-time — i.e. a policy whose deny\u002Faudit decision depends on data from elsewhere in the subscription (sibling\u002Fparent resource state, RG-wide invariants, multi-hop relationships, etc.). Formally called Azure Policy External Evaluation; sometimes referred to colloquially as \"Invoke\". Drives an iterative KQL co-design loop against the user's real subscription via `az graph query`, then emits a policy definition, assignment, `.http` test flow, and an `EXPLANATION.md` companion. Read-only; never provisions anything.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1073,1074,1077,1080],{"name":11,"slug":8,"type":16},{"name":1075,"slug":1076,"type":16},"Compliance","compliance",{"name":1078,"slug":1079,"type":16},"Governance","governance",{"name":1081,"slug":1082,"type":16},"Policy","policy",1686,"https:\u002F\u002Fgithub.com\u002FAzure\u002Fazure-policy","2026-07-12T08:17:48.378432",{"slug":1087,"name":1087,"fn":1088,"description":1089,"org":1090,"tags":1091,"stars":1102,"repoUrl":1103,"updatedAt":1104},"azure-blueprints-migration","migrate Azure Blueprints to Template Specs","Use when a user needs to migrate off Azure Blueprints (definitions and\u002For assignments) to Template Specs and Deployment Stacks before the January 31, 2027 retirement. Covers inventory, export, conversion to Bicep, policy decoupling, Template Spec publishing, Deployment Stack deployment with deny-settings, validation, and cutover.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1092,1093,1096,1099],{"name":11,"slug":8,"type":16},{"name":1094,"slug":1095,"type":16},"Deployment","deployment",{"name":1097,"slug":1098,"type":16},"Infrastructure as Code","infrastructure-as-code",{"name":1100,"slug":1101,"type":16},"Migration","migration",260,"https:\u002F\u002Fgithub.com\u002FAzure\u002Fazure-blueprints","2026-07-12T08:17:49.646405",{"slug":1106,"name":1106,"fn":1107,"description":1108,"org":1109,"tags":1110,"stars":1121,"repoUrl":1122,"updatedAt":1123},"apiview-feedback-resolution","resolve APIView feedback on Azure SDKs","Analyze and resolve APIView review feedback on Azure SDK PRs. **UTILITY SKILL**. USE FOR: APIView comments, API review feedback, SDK API surface changes. DO NOT USE FOR: general code review, non-APIView feedback. INVOKES: azure-sdk-mcp:azsdk_apiview_get_comments, azure-sdk-mcp:azsdk_typespec_customized_code_update.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1111,1114,1115,1118],{"name":1112,"slug":1113,"type":16},"API Development","api-development",{"name":11,"slug":8,"type":16},{"name":1116,"slug":1117,"type":16},"Code Review","code-review",{"name":1119,"slug":1120,"type":16},"Documentation","documentation",133,"https:\u002F\u002Fgithub.com\u002FAzure\u002Fazure-sdk-tools","2026-07-12T08:17:43.350876",{"slug":1125,"name":1125,"fn":1126,"description":1127,"org":1128,"tags":1129,"stars":1121,"repoUrl":1122,"updatedAt":1138},"azsdk-common-live-and-recorded-tests","deploy resources and run Azure SDK tests","Deploy test resources and run Azure SDK tests in live, record, or playback mode. WHEN: \"run live tests\", \"run recorded tests\", \"deploy test resources\", \"record tests\", \"run tests in record mode\", \"clean up test resources\", \"run tests against live resources\". DO NOT USE FOR: writing new tests, authoring Bicep templates, playback-only test runs without resource deployment. INVOKES: azure-sdk-mcp:azsdk_package_run_tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1130,1131,1132,1135],{"name":11,"slug":8,"type":16},{"name":1094,"slug":1095,"type":16},{"name":1133,"slug":1134,"type":16},"SDK","sdk",{"name":1136,"slug":1137,"type":16},"Testing","testing","2026-07-12T08:17:44.718943",{"slug":1140,"name":1140,"fn":1141,"description":1142,"org":1143,"tags":1144,"stars":1121,"repoUrl":1122,"updatedAt":1153},"azsdk-common-prepare-release-plan","manage Azure SDK release plan work items","Create, get, update, abandon, and link SDK PRs to release plan work items for Azure SDK releases. **UTILITY SKILL**. USE FOR: \"create release plan\", \"get release plan\", \"update release plan\", \"update API spec in release plan\", \"update SDK details in release plan\", \"abandon release plan\", \"link SDK PR to plan\", \"namespace approval\", \"check release plan status\". DO NOT USE FOR: SDK code generation, pipeline troubleshooting, API review feedback. INVOKES: azure-sdk-mcp:azsdk_create_release_plan, azure-sdk-mcp:azsdk_get_release_plan, azure-sdk-mcp:azsdk_get_release_plan_for_spec_pr, azure-sdk-mcp:azsdk_update_release_plan, azure-sdk-mcp:azsdk_update_api_spec_pull_request_in_release_plan, azure-sdk-mcp:azsdk_update_sdk_details_in_release_plan, azure-sdk-mcp:azsdk_abandon_release_plan, azure-sdk-mcp:azsdk_link_sdk_pull_request_to_release_plan, azure-sdk-mcp:azsdk_link_namespace_approval_issue.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1145,1146,1149,1152],{"name":11,"slug":8,"type":16},{"name":1147,"slug":1148,"type":16},"GitHub","github",{"name":1150,"slug":1151,"type":16},"Project Management","project-management",{"name":1133,"slug":1134,"type":16},"2026-07-12T08:17:38.345387",{"slug":1155,"name":1155,"fn":1156,"description":1157,"org":1158,"tags":1159,"stars":1121,"repoUrl":1122,"updatedAt":1166},"azsdk-common-sdk-release","release Azure SDK packages","Check release readiness and trigger the release pipeline for Azure SDK packages. **UTILITY SKILL**. USE FOR: \"release SDK\", \"trigger release\", \"check release readiness\", \"release pipeline\", \"publish package\", \"ship SDK\". DO NOT USE FOR: SDK development, code generation, pipeline debugging, release plan creation. INVOKES: azure-sdk-mcp:azsdk_release_sdk.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1160,1161,1164,1165],{"name":11,"slug":8,"type":16},{"name":1162,"slug":1163,"type":16},"CI\u002FCD","ci-cd",{"name":1094,"slug":1095,"type":16},{"name":1133,"slug":1134,"type":16},"2026-07-12T08:17:34.27607",{"slug":1168,"name":1168,"fn":1169,"description":1170,"org":1171,"tags":1172,"stars":1121,"repoUrl":1122,"updatedAt":1181},"azure-typespec-author","author and modify Azure TypeSpec API specifications","Authors and modifies Azure TypeSpec (.tsp) API specifications. USE FOR: any TypeSpec\u002Ftsp change — api versions (add, bump, preview, stable, promote), resources, operations, models, properties, decorators, visibility, constraints, breaking changes, LRO, suppressions, operationId, spread model. Covers ARM resource-manager and data-plane services. DO NOT USE FOR: SDK generation, releasing SDK packages, or single MCP tool calls. INVOKES: azure-sdk-mcp:azsdk_typespec_generate_authoring_plan, azure-sdk-mcp:azsdk_run_typespec_validation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1173,1174,1175,1178],{"name":1112,"slug":1113,"type":16},{"name":11,"slug":8,"type":16},{"name":1176,"slug":1177,"type":16},"OpenAPI","openapi",{"name":1179,"slug":1180,"type":16},"Technical Writing","technical-writing","2026-07-12T08:17:39.603232",{"slug":1183,"name":1183,"fn":1184,"description":1185,"org":1186,"tags":1187,"stars":1121,"repoUrl":1122,"updatedAt":1192},"generate-sdk-locally","generate and test Azure SDKs locally","Generate, build, and test Azure SDKs locally from TypeSpec with automatic customization. WHEN: \"generate SDK locally\", \"build SDK\", \"run SDK tests\", \"run CI checks\", \"validate package\", \"run checks\", \"update changelog\", \"fix SDK build errors\", \"fix breaking changes\", \"resolve SDK generation errors\", \"customize TypeSpec\", \"rename SDK client\", \"rename SDK model\", \"hide operation from SDK\", \"fix analyzer errors\", \"resolve customization drift\", \"create subclient\", \"update metadata\", \"update version\". DO NOT USE FOR: publishing to package registries, CI pipeline configuration, API design review. INVOKES: azsdk_verify_setup, azsdk_package_generate_code, azsdk_package_build_code, azsdk_package_run_check, azsdk_package_run_tests, azsdk_customized_code_update, azsdk_package_update_changelog_content, azsdk_package_update_metadata, azsdk_package_update_version.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1188,1189,1190,1191],{"name":11,"slug":8,"type":16},{"name":1162,"slug":1163,"type":16},{"name":1133,"slug":1134,"type":16},{"name":1136,"slug":1137,"type":16},"2026-07-12T08:17:37.08523",{"slug":1194,"name":1194,"fn":1195,"description":1196,"org":1197,"tags":1198,"stars":1121,"repoUrl":1122,"updatedAt":1206},"markdown-token-optimizer","optimize markdown files for token efficiency","Analyze markdown files for token efficiency and reduce context-window bloat. **UTILITY SKILL**. DO NOT USE FOR: code optimization, general file editing, non-markdown files. TRIGGERS: optimize markdown, reduce tokens, token count, token bloat, too many tokens, make concise, shrink file, file too large, optimize for AI, token efficiency, verbose markdown, reduce file size. INVOKES: waza CLI.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1199,1202,1205],{"name":1200,"slug":1201,"type":16},"LLM","llm",{"name":1203,"slug":1204,"type":16},"Performance","performance",{"name":1179,"slug":1180,"type":16},"2026-07-12T08:17:42.080413",{"slug":1208,"name":1208,"fn":1209,"description":1210,"org":1211,"tags":1212,"stars":1121,"repoUrl":1122,"updatedAt":1219},"pipeline-troubleshooting","troubleshoot Azure SDK CI pipelines","Diagnose and resolve failures in Azure SDK CI and generation pipelines. **UTILITY SKILL**. USE FOR: \"pipeline failed\", \"build failure\", \"CI check failing\", \"SDK generation error\", \"reproduce pipeline locally\", \"debug SDK pipeline\". DO NOT USE FOR: local build issues without pipeline context, API design review, SDK publishing. INVOKES: azure-sdk-mcp:azsdk_analyze_pipeline, azure-sdk-mcp:azsdk_package_build_code, azure-sdk-mcp:azsdk_package_run_check.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1213,1214,1215,1218],{"name":11,"slug":8,"type":16},{"name":1162,"slug":1163,"type":16},{"name":1216,"slug":1217,"type":16},"Debugging","debugging",{"name":1133,"slug":1134,"type":16},"2026-07-12T08:17:40.821512",{"slug":1221,"name":1221,"fn":1222,"description":1223,"org":1224,"tags":1225,"stars":1121,"repoUrl":1122,"updatedAt":1232},"sensei","improve skill frontmatter compliance","**WORKFLOW SKILL** — Iteratively improve skill frontmatter compliance using the Ralph loop pattern. WHEN: \"run sensei\", \"sensei help\", \"improve skill\", \"fix frontmatter\", \"skill compliance\", \"frontmatter audit\", \"score skill\", \"check skill tokens\". INVOKES: token counting tools, test runners, git commands. FOR SINGLE OPERATIONS: use token CLI directly for counts\u002Fchecks.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1226,1227,1228,1231],{"name":11,"slug":8,"type":16},{"name":1075,"slug":1076,"type":16},{"name":1229,"slug":1230,"type":16},"Process Optimization","process-optimization",{"name":1179,"slug":1180,"type":16},"2026-07-12T08:17:32.970921",{"slug":1234,"name":1234,"fn":1235,"description":1236,"org":1237,"tags":1238,"stars":1121,"repoUrl":1122,"updatedAt":1244},"skill-authoring","author agent skills for agentskills.io","Write Agent Skills that comply with the agentskills.io specification. WHEN: \"create a skill\", \"new skill\", \"write a skill\", \"skill template\", \"skill structure\", \"review skill\", \"skill PR\", \"skill compliance\", \"SKILL.md format\", \"skill frontmatter\", \"skill best practices\". DO NOT USE FOR: improving existing skills (use sensei), general documentation. INVOKES: waza CLI.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":11},[1239,1240,1243],{"name":1119,"slug":1120,"type":16},{"name":1241,"slug":1242,"type":16},"Plugin Development","plugin-development",{"name":1179,"slug":1180,"type":16},"2026-07-12T08:17:35.873862",109]