[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-doca-telemetry-utils":3,"mdc-yh52nf-key":31,"related-repo-nvidia-doca-telemetry-utils":1462,"related-org-nvidia-doca-telemetry-utils":1568},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":26,"sourceUrl":29,"mdContent":30},"doca-telemetry-utils","analyze DOCA telemetry and diagnostic counters","Use this skill when the user is invoking `doca_telemetry_utils` on a host with DOCA installed — discovering the diagnostic-counter schema, translating counter names to binary Data IDs, validating per-device counter support before committing a DOCA Telemetry exporter config, or reverse-resolving a captured Data ID. Trigger even when the user does not explicitly mention \"doca_telemetry_utils\" or \"Data ID\" — typical implicit phrasings include \"my exporter ships but the collector sees nothing\", \"this metric silently drops downstream\", \"which counters does this BlueField expose\", \"translate this 0x... back to a counter name\", \"what do node \u002F pcie_index \u002F depth mean here\", or \"is this counter supported on this device before I commit it\". Refuse and route elsewhere for developer-side collector \u002F exporter library programming, DTS deployment, or DOCA install \u002F repair — those belong to doca-telemetry, doca-public-knowledge-map, and doca-setup.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19],{"name":13,"slug":14,"type":15},"Observability","observability","tag",{"name":17,"slug":18,"type":15},"Monitoring","monitoring",{"name":9,"slug":8,"type":15},2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-20T06:24:16.350074","Apache-2.0",281,[],{"repoUrl":21,"stars":20,"forks":24,"topics":27,"description":28},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fdoca-telemetry-utils","---\nlicense: Apache-2.0\nname: doca-telemetry-utils\ndescription: >\n  Use this skill when the user is invoking `doca_telemetry_utils` on\n  a host with DOCA installed — discovering the diagnostic-counter\n  schema, translating counter names to binary Data IDs, validating\n  per-device counter support before committing a DOCA Telemetry\n  exporter config, or reverse-resolving a captured Data ID. Trigger\n  even when the user does not explicitly mention \"doca_telemetry_utils\"\n  or \"Data ID\" — typical implicit phrasings include \"my exporter ships\n  but the collector sees nothing\", \"this metric silently drops\n  downstream\", \"which counters does this BlueField expose\", \"translate\n  this 0x... back to a counter name\", \"what do node \u002F pcie_index \u002F\n  depth mean here\", or \"is this counter supported on this device before\n  I commit it\". Refuse and route elsewhere for developer-side collector\n  \u002F exporter library programming, DTS deployment, or DOCA install \u002F\n  repair — those belong to doca-telemetry, doca-public-knowledge-map,\n  and doca-setup.\nmetadata:\n  kind: tool\ncompatibility: >\n  Requires DOCA SDK installed at \u002Fopt\u002Fmellanox\u002Fdoca on Linux (Ubuntu\n  22.04\u002F24.04 or RHEL\u002FSLES) with the Telemetry optional component and\n  a BlueField DPU visible to DOCA on a known PCI address. Invokes\n  \u002Fopt\u002Fmellanox\u002Fdoca\u002Ftools\u002Fdoca_telemetry_utils; per-device probe\n  typically requires elevated privileges.\n---\n\n# DOCA Telemetry Utils\n\n**Where to start:** This is a tool skill for invoking\n`doca_telemetry_utils` — the documented host-side CLI that\nsupports a DOCA Telemetry exporter \u002F collector pipeline by\ndiscovering the counter schema, translating counter names ↔\nData IDs, and probing per-device counter support. Open\n[`TASKS.md`](TASKS.md) and start at\n[`## install`](TASKS.md#install) for the host-side\nprerequisites and [`## run`](TASKS.md#run) for the\nthree documented invocation classes\n(enumerate \u002F name→ID \u002F ID→name). Open\n[`CAPABILITIES.md`](CAPABILITIES.md) when the question is\n*what does this tool actually discover about the telemetry\nschema*, *how does it pair with the developer-side\n[`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md) and\nexporter libraries*, *how do I confirm a device supports a\ncounter before committing an exporter config to it*, or\n*why does my exporter pipeline silently drop a metric*.\n\nThis skill is the **operator-side support tool** for a\nDOCA Telemetry deployment. It is NOT the developer-side\ncollector library (that is\n[`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md)),\nNOT the developer-side publisher library (that is\n`doca-telemetry-exporter` — see\n[`doca-telemetry ## Related skills`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md#related-skills)),\nand NOT a DOCA Telemetry Service (DTS) deployment guide\n(route via\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)).\nThree separate surfaces; conflating them is the most\ncommon telemetry first-touch error.\n\n## Example questions this skill answers well\n\nThe CLASSES of `doca_telemetry_utils` questions this skill\nis built to answer, each with one worked example. The\nclass is the load-bearing piece; the worked example is one\ninstance.\n\n- **\"My exporter says it's emitting `port_rx_bytes` but\n  nothing shows up downstream — what did I get wrong?\"** —\n  worked example: *\"my exporter config has a counter name\n  string and my collector sees no events with that\n  name\"*. Answered by the name ↔ Data ID translation\n  step in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the per-device-support probe in\n  [`TASKS.md ## test`](TASKS.md#test): the exporter\n  ships a Data ID, not a name; a name in the config that\n  resolves to a Data ID the device does not support is\n  silently dropped.\n- **\"Which DOCA diagnostic counters does this BlueField\n  actually expose?\"** — worked example: *\"enumerate the\n  full counter schema for my BlueField-3 before I write\n  the exporter config\"*. Answered by the schema-discovery\n  invocation class in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  (`doca_telemetry_utils get-counters` lists every\n  counter name the diagnostic-data surface knows about;\n  pair with a per-device probe to confirm support).\n- **\"I have a Data ID in a captured log — what counter\n  was that?\"** — worked example: *\"a downstream consumer\n  emitted `Data ID: 0x1160000600030201` — translate it\n  back so I can correlate against the public guide\"*.\n  Answered by the reverse-resolve invocation class in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the\n  ID-encodes-properties rule in\n  [`TASKS.md ## use`](TASKS.md#use) (a Data ID carries\n  the counter's property dimensions; the reverse-resolve\n  reports them).\n- **\"What property dimensions does this counter take and\n  what values are valid?\"** — worked example: *\"I know\n  the counter is `pcie_link_write_stalled_time_*` — what\n  do `node` \u002F `pcie_index` \u002F `depth` mean and what\n  values does the device accept?\"*. Answered by the\n  property-dimension table in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the `\u003Cname>` invocation without arguments which\n  prints the documented property options + units +\n  unit-specific axes.\n- **\"Is this counter supported on this device before I\n  commit it to the exporter config?\"** — worked example:\n  *\"validate that `port_rx_bytes` with `node=1` is\n  exposed on the BlueField at PCIe address X before I\n  write the exporter config\"*. Answered by the\n  per-device-support probe (`\u003Cdevice PCI> \u003Cname>\n  [properties]`) in\n  [`CAPABILITIES.md ## Capabilities and modes`](CAPABILITIES.md#capabilities-and-modes)\n  + the gate-before-commit rule in\n  [`TASKS.md ## use`](TASKS.md#use).\n- **\"Is `doca_telemetry_utils` on my install, and is it\n  paired with the matching `doca-telemetry` library\n  version?\"** — worked example: *\"is the diagnostic-\n  data counter set my exporter targets on this DOCA\n  version?\"*. Answered by the version-overlay in\n  [`CAPABILITIES.md ## Version compatibility`](CAPABILITIES.md#version-compatibility),\n  which redirects to the canonical\n  [`doca-version`](..\u002F..\u002Fdoca-version\u002FSKILL.md) chain\n  and adds the *tool ↔ `doca-telemetry` library\n  schema-version* match rule.\n\n## Audience\n\nThis skill serves **external operators, developers, and\nAI agents standing up or debugging a DOCA Telemetry\nexporter \u002F collector pipeline who need to confirm the\ncounter schema, validate per-device support, or\ntranslate between human-readable counter names and the\nbinary Data IDs the exporter actually ships**.\nConcretely:\n\n- A platform operator standing up a new DOCA Telemetry\n  exporter on a BlueField fleet who needs to confirm\n  which counters the target devices actually expose\n  before committing the exporter config.\n- A developer of a downstream consumer (a collector\n  app linking [`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md),\n  or a third-party aggregator consuming via DTS) who\n  has a captured Data ID stream and needs to translate\n  IDs back to counter names + properties.\n- An operator debugging a *\"nothing is shipping\n  downstream\"* \u002F *\"this metric is silently missing\"*\n  report against a deployed exporter — the\n  schema-discovery + per-device-support probes are the\n  canonical *\"is the counter even supposed to work on\n  this device\"* first step before suspecting the\n  collector or the network path.\n- An AI agent producing a *\"validated exporter config\n  for this BlueField + this DOCA version\"* answer\n  honestly — with each counter resolved to its Data ID\n  and each Data ID confirmed against the per-device\n  capability probe.\n\nIt is **not** for users debugging the `doca_telemetry_utils`\nbinary itself, **not** a substitute for the live public\nDOCA Telemetry guides, and **not** the right place for\nlearning how to *write* an exporter application (that\naudience belongs in\n[`doca-telemetry-exporter` skill](..\u002F..\u002Flibs\u002Fdoca-telemetry-exporter\u002FSKILL.md)\nwhen present, or via\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)).\n\nThe tool is shipped as a **CLI binary** under\n`\u002Fopt\u002Fmellanox\u002Fdoca\u002Ftools\u002F`, not a library you link\nagainst. The skill uses the same `kind: tool`\nthree-file shape as the rest of the bundle so the\nagent's task-verb contract is uniform across libraries,\nservices, and tools.\n\n## Language scope\n\n`doca_telemetry_utils` is a C host-side CLI that uses\nthe documented `doca_telemetry_diag` surface to list\ncounter types and probe per-device support. Its inputs\nare command-line arguments (counter name, property\nvalues, Data ID, optional device PCI address); its\noutputs are human-readable Data ID + properties\nmappings. The skill keeps workflow guidance language-\nneutral; downstream consumers in any language can use\nthe resolved Data IDs against their own collector code.\n\n## When to load this skill\n\nLoad this skill when the user is — or the agent needs\nto — invoke `doca_telemetry_utils` on a real host with\nDOCA installed, alongside an exporter \u002F collector\npipeline that needs schema discovery, per-device\nsupport validation, or Data ID translation.\nConcretely:\n\n- Enumerating the full counter schema before writing\n  a fresh DOCA Telemetry exporter config (`get-counters`).\n- Validating that a chosen counter + property set\n  resolves to a Data ID the target device actually\n  supports, before committing the config.\n- Reverse-resolving a Data ID captured from a\n  downstream consumer back to a counter name +\n  properties for correlation against the public guide.\n- Debugging a *\"exporter ships, collector receives\n  nothing\"* report (the canonical schema-mismatch \u002F\n  unsupported-counter failure mode).\n- Producing a *\"validated counter set for this\n  BlueField + this DOCA version\"* artifact as part of\n  a structured exporter-config baseline.\n- Migrating an exporter pipeline across DOCA versions\n  and confirming each previously-supported counter\n  still resolves cleanly on the new version.\n\nDo **not** load this skill for general DOCA orientation,\ncollector \u002F exporter library programming, DTS\ndeployment, or DOCA install. For those, route to\n[`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md),\n[`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md),\nor [`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md).\n\n## What this skill provides\n\nThis is a **thin loader**. Substantive material lives\nin two companion files:\n\n- `CAPABILITIES.md` — what `doca_telemetry_utils`\n  discovers + how it pairs with the developer-side\n  surfaces: the three documented invocation classes\n  (enumerate \u002F name→ID \u002F ID→name), the property-\n  dimension model (counters carry property axes such\n  as `node`, `pcie_index`, `depth`, plus per-unit\n  axes), the optional per-device capability probe\n  (`\u003Cdevice PCI> \u003Cname>` runs the resolved counter\n  against the device), the operator-side support\n  role (this is NOT the developer-side library; it\n  exists to make exporter \u002F collector setups\n  honest), the version overlay (tool ↔\n  `doca-telemetry` library schema version pairing),\n  the layered error taxonomy (install \u002F parse \u002F\n  unknown-counter \u002F unknown-data-id \u002F property-out-\n  of-range \u002F device-not-supported \u002F version \u002F\n  cross-cutting), the observability surface\n  (stdout-only), and the safety policy (the tool is\n  read-only; mistakes appear downstream as silent\n  metric drops, not crashes).\n- `TASKS.md` — step-by-step workflows for the\n  in-scope task verbs: `install` (host-side DOCA +\n  telemetry component prerequisites), `configure`\n  (axis decisions: which invocation class, which\n  device, which counter), `build` (route to install\n  — the binary is shipped), `modify` (refuse — do\n  not patch the binary; modify the invocation and\n  the exporter \u002F collector config that consumes the\n  resolved Data IDs), `run` (the three documented\n  invocations), `test` (round-trip a chosen\n  counter through name → Data ID → per-device\n  probe → exporter config → collector receipt),\n  `debug` (walk the error taxonomy), `use` (the\n  hand-off into the developer-side `doca-telemetry`\n  collector \u002F exporter pipeline), plus a `Deferred\n  task verbs` block.\n\nThe skill assumes a host where DOCA is already\ninstalled with the telemetry component, a BlueField\nthe operator is targeting is visible to DOCA, and the\nexporter \u002F collector pipeline that consumes the\nresolved Data IDs is either being authored or has\nalready been authored against the matching\n[`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md)\nlibrary version.\n\n## What this skill deliberately does not ship\n\nThis skill is **agent guidance**, not a samples or\nscripts bundle. To keep the boundary clean, it\ndeliberately does not contain — and pull requests\nshould not add:\n\n- **A verbatim counter inventory or Data ID\n  table.** The counter set evolves across DOCA\n  versions and BlueField generations; an\n  inventory pinned in this skill would silently rot.\n  `doca_telemetry_utils get-counters` on the\n  installed binary is the authoritative source.\n- **Pre-baked exporter \u002F collector configs.** Configs\n  are install-, device-, and use-case-specific; a\n  packaged config in this skill would mislead\n  operators on a different setup.\n- **A DTS deployment recipe.** DTS is a separate DOCA\n  service with its own public guide; route via\n  [`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md).\n  This skill *resolves the counter names + IDs* a\n  DTS pipeline consumes; it does not configure DTS.\n- **Wrappers, parsers, or scripts** in any language\n  that consume the tool's output. The output is a\n  simple text mapping; users who want to script\n  against it should read the live guide and write\n  the parser against their installed version.\n- **A `samples\u002F` or `reference\u002F` subtree.** This is\n  a thin loader for a shipped CLI; substantive\n  material lives on the public page, in `--help`,\n  and in [`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md).\n\n## Loading order\n\n1. Read this `SKILL.md` first to confirm the user's\n   question is in scope (operator-side support\n   tooling for a telemetry pipeline, not the\n   developer-side library programming).\n2. **For the three invocation classes, the\n   property-dimension model, the per-device support\n   probe, the version overlay, the error taxonomy,\n   the observability surface, and the safety\n   policy, see [CAPABILITIES.md](CAPABILITIES.md).**\n3. **For the documented invocations and the\n   discover → resolve → validate → consume\n   workflow — `install`, `configure`, `build`,\n   `modify`, `run`, `test`, `debug`, `use` — see\n   [TASKS.md](TASKS.md).**\n\n## Related skills\n\n- [`doca-telemetry`](..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md)\n  — the developer-side collector library whose\n  schema this tool helps the operator discover. Pair\n  them in every exporter-pipeline triage session.\n  The collector library skill teaches the\n  collector-vs-exporter rule, the schema-must-match\n  contract with the publisher, and the consumer-\n  queue-full back-pressure rule; this tool's role\n  is to make the *schema* half of that contract\n  inspectable from the operator side. Conflating\n  the library and the tool is the most common\n  telemetry first-touch error.\n- [`doca-public-knowledge-map`](..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md)\n  — routing to the public DOCA Telemetry guide and\n  the DOCA Telemetry Service (DTS) page on\n  `docs.nvidia.com`, plus the on-disk install\n  layout for the tool.\n- [`doca-version`](..\u002F..\u002Fdoca-version\u002FSKILL.md) —\n  canonical DOCA version-handling rules. The\n  `## Version compatibility` section in\n  [`CAPABILITIES.md`](CAPABILITIES.md) is a concise\n  overlay that redirects here for the body and\n  adds the *tool ↔ `doca-telemetry` library\n  schema-version pairing* rule.\n- [`doca-setup`](..\u002F..\u002Fdoca-setup\u002FSKILL.md) — env\n  preparation, install verification, and the *I\n  have no install yet* path with the public NGC\n  DOCA container. This skill assumes its\n  preconditions are satisfied (DOCA installed\n  with the telemetry component, BlueField visible\n  to DOCA).\n- [`doca-debug`](..\u002F..\u002Fdoca-debug\u002FSKILL.md) — the\n  cross-cutting debug ladder. Telemetry-utils\n  feeds the cross-cutting ladder by surfacing the\n  counter schema + per-device support truth at the\n  runtime layer; exporter-pipeline regressions\n  often resolve at the schema layer before\n  touching the collector \u002F network path.\n- [`doca-structured-tools-contract`](..\u002F..\u002Fdoca-structured-tools-contract\u002FSKILL.md)\n  — the bundle's detect → prefer → fall back →\n  report contract for structured helper tools. The\n  command appendix in [`TASKS.md`](TASKS.md)\n  honors this contract.\n- [`doca-programming-guide`](..\u002F..\u002Fdoca-programming-guide\u002FSKILL.md)\n  — general DOCA programming patterns shared by\n  every library \u002F tool surface, including the\n  cross-library `DOCA_ERROR_*` taxonomy this\n  tool's host-side error layer overlays on top of\n  when downstream collector \u002F exporter code\n  surfaces a related error.\n",{"data":32,"body":36},{"license":23,"name":4,"description":6,"metadata":33,"compatibility":35},{"kind":34},"tool","Requires DOCA SDK installed at \u002Fopt\u002Fmellanox\u002Fdoca on Linux (Ubuntu 22.04\u002F24.04 or RHEL\u002FSLES) with the Telemetry optional component and a BlueField DPU visible to DOCA on a known PCI address. Invokes \u002Fopt\u002Fmellanox\u002Fdoca\u002Ftools\u002Fdoca_telemetry_utils; per-device probe typically requires elevated privileges.\n",{"type":37,"children":38},"root",[39,47,155,209,216,228,579,585,597,657,718,746,752,770,776,788,842,884,890,902,1045,1060,1066,1078,1188,1194,1285,1291],{"type":40,"tag":41,"props":42,"children":43},"element","h1",{"id":4},[44],{"type":45,"value":46},"text","DOCA Telemetry Utils",{"type":40,"tag":48,"props":49,"children":50},"p",{},[51,57,59,66,68,78,80,90,92,102,104,113,115,121,123,140,141,146,148,153],{"type":40,"tag":52,"props":53,"children":54},"strong",{},[55],{"type":45,"value":56},"Where to start:",{"type":45,"value":58}," This is a tool skill for invoking\n",{"type":40,"tag":60,"props":61,"children":63},"code",{"className":62},[],[64],{"type":45,"value":65},"doca_telemetry_utils",{"type":45,"value":67}," — the documented host-side CLI that\nsupports a DOCA Telemetry exporter \u002F collector pipeline by\ndiscovering the counter schema, translating counter names ↔\nData IDs, and probing per-device counter support. Open\n",{"type":40,"tag":69,"props":70,"children":72},"a",{"href":71},"TASKS.md",[73],{"type":40,"tag":60,"props":74,"children":76},{"className":75},[],[77],{"type":45,"value":71},{"type":45,"value":79}," and start at\n",{"type":40,"tag":69,"props":81,"children":83},{"href":82},"TASKS.md#install",[84],{"type":40,"tag":60,"props":85,"children":87},{"className":86},[],[88],{"type":45,"value":89},"## install",{"type":45,"value":91}," for the host-side\nprerequisites and ",{"type":40,"tag":69,"props":93,"children":95},{"href":94},"TASKS.md#run",[96],{"type":40,"tag":60,"props":97,"children":99},{"className":98},[],[100],{"type":45,"value":101},"## run",{"type":45,"value":103}," for the\nthree documented invocation classes\n(enumerate \u002F name→ID \u002F ID→name). Open\n",{"type":40,"tag":69,"props":105,"children":107},{"href":106},"CAPABILITIES.md",[108],{"type":40,"tag":60,"props":109,"children":111},{"className":110},[],[112],{"type":45,"value":106},{"type":45,"value":114}," when the question is\n",{"type":40,"tag":116,"props":117,"children":118},"em",{},[119],{"type":45,"value":120},"what does this tool actually discover about the telemetry\nschema",{"type":45,"value":122},", ",{"type":40,"tag":116,"props":124,"children":125},{},[126,128,138],{"type":45,"value":127},"how does it pair with the developer-side\n",{"type":40,"tag":69,"props":129,"children":131},{"href":130},"..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md",[132],{"type":40,"tag":60,"props":133,"children":135},{"className":134},[],[136],{"type":45,"value":137},"doca-telemetry",{"type":45,"value":139}," and\nexporter libraries",{"type":45,"value":122},{"type":40,"tag":116,"props":142,"children":143},{},[144],{"type":45,"value":145},"how do I confirm a device supports a\ncounter before committing an exporter config to it",{"type":45,"value":147},", or\n",{"type":40,"tag":116,"props":149,"children":150},{},[151],{"type":45,"value":152},"why does my exporter pipeline silently drop a metric",{"type":45,"value":154},".",{"type":40,"tag":48,"props":156,"children":157},{},[158,160,165,167,175,177,183,185,195,197,207],{"type":45,"value":159},"This skill is the ",{"type":40,"tag":52,"props":161,"children":162},{},[163],{"type":45,"value":164},"operator-side support tool",{"type":45,"value":166}," for a\nDOCA Telemetry deployment. It is NOT the developer-side\ncollector library (that is\n",{"type":40,"tag":69,"props":168,"children":169},{"href":130},[170],{"type":40,"tag":60,"props":171,"children":173},{"className":172},[],[174],{"type":45,"value":137},{"type":45,"value":176},"),\nNOT the developer-side publisher library (that is\n",{"type":40,"tag":60,"props":178,"children":180},{"className":179},[],[181],{"type":45,"value":182},"doca-telemetry-exporter",{"type":45,"value":184}," — see\n",{"type":40,"tag":69,"props":186,"children":188},{"href":187},"..\u002F..\u002Flibs\u002Fdoca-telemetry\u002FSKILL.md#related-skills",[189],{"type":40,"tag":60,"props":190,"children":192},{"className":191},[],[193],{"type":45,"value":194},"doca-telemetry ## Related skills",{"type":45,"value":196},"),\nand NOT a DOCA Telemetry Service (DTS) deployment guide\n(route via\n",{"type":40,"tag":69,"props":198,"children":200},{"href":199},"..\u002F..\u002Fdoca-public-knowledge-map\u002FSKILL.md",[201],{"type":40,"tag":60,"props":202,"children":204},{"className":203},[],[205],{"type":45,"value":206},"doca-public-knowledge-map",{"type":45,"value":208},").\nThree separate surfaces; conflating them is the most\ncommon telemetry first-touch error.",{"type":40,"tag":210,"props":211,"children":213},"h2",{"id":212},"example-questions-this-skill-answers-well",[214],{"type":45,"value":215},"Example questions this skill answers well",{"type":40,"tag":48,"props":217,"children":218},{},[219,221,226],{"type":45,"value":220},"The CLASSES of ",{"type":40,"tag":60,"props":222,"children":224},{"className":223},[],[225],{"type":45,"value":65},{"type":45,"value":227}," questions this skill\nis built to answer, each with one worked example. The\nclass is the load-bearing piece; the worked example is one\ninstance.",{"type":40,"tag":229,"props":230,"children":231},"ul",{},[232,288,323,375,446,511],{"type":40,"tag":233,"props":234,"children":235},"li",{},[236,249,251,256,258,268],{"type":40,"tag":52,"props":237,"children":238},{},[239,241,247],{"type":45,"value":240},"\"My exporter says it's emitting ",{"type":40,"tag":60,"props":242,"children":244},{"className":243},[],[245],{"type":45,"value":246},"port_rx_bytes",{"type":45,"value":248}," but\nnothing shows up downstream — what did I get wrong?\"",{"type":45,"value":250}," —\nworked example: ",{"type":40,"tag":116,"props":252,"children":253},{},[254],{"type":45,"value":255},"\"my exporter config has a counter name\nstring and my collector sees no events with that\nname\"",{"type":45,"value":257},". Answered by the name ↔ Data ID translation\nstep in\n",{"type":40,"tag":69,"props":259,"children":261},{"href":260},"CAPABILITIES.md#capabilities-and-modes",[262],{"type":40,"tag":60,"props":263,"children":265},{"className":264},[],[266],{"type":45,"value":267},"CAPABILITIES.md ## Capabilities and modes",{"type":40,"tag":229,"props":269,"children":270},{},[271],{"type":40,"tag":233,"props":272,"children":273},{},[274,276,286],{"type":45,"value":275},"the per-device-support probe in\n",{"type":40,"tag":69,"props":277,"children":279},{"href":278},"TASKS.md#test",[280],{"type":40,"tag":60,"props":281,"children":283},{"className":282},[],[284],{"type":45,"value":285},"TASKS.md ## test",{"type":45,"value":287},": the exporter\nships a Data ID, not a name; a name in the config that\nresolves to a Data ID the device does not support is\nsilently dropped.",{"type":40,"tag":233,"props":289,"children":290},{},[291,296,298,303,305,313,315,321],{"type":40,"tag":52,"props":292,"children":293},{},[294],{"type":45,"value":295},"\"Which DOCA diagnostic counters does this BlueField\nactually expose?\"",{"type":45,"value":297}," — worked example: ",{"type":40,"tag":116,"props":299,"children":300},{},[301],{"type":45,"value":302},"\"enumerate the\nfull counter schema for my BlueField-3 before I write\nthe exporter config\"",{"type":45,"value":304},". Answered by the schema-discovery\ninvocation class in\n",{"type":40,"tag":69,"props":306,"children":307},{"href":260},[308],{"type":40,"tag":60,"props":309,"children":311},{"className":310},[],[312],{"type":45,"value":267},{"type":45,"value":314},"\n(",{"type":40,"tag":60,"props":316,"children":318},{"className":317},[],[319],{"type":45,"value":320},"doca_telemetry_utils get-counters",{"type":45,"value":322}," lists every\ncounter name the diagnostic-data surface knows about;\npair with a per-device probe to confirm support).",{"type":40,"tag":233,"props":324,"children":325},{},[326,331,332,345,347,355],{"type":40,"tag":52,"props":327,"children":328},{},[329],{"type":45,"value":330},"\"I have a Data ID in a captured log — what counter\nwas that?\"",{"type":45,"value":297},{"type":40,"tag":116,"props":333,"children":334},{},[335,337,343],{"type":45,"value":336},"\"a downstream consumer\nemitted ",{"type":40,"tag":60,"props":338,"children":340},{"className":339},[],[341],{"type":45,"value":342},"Data ID: 0x1160000600030201",{"type":45,"value":344}," — translate it\nback so I can correlate against the public guide\"",{"type":45,"value":346},".\nAnswered by the reverse-resolve invocation class in\n",{"type":40,"tag":69,"props":348,"children":349},{"href":260},[350],{"type":40,"tag":60,"props":351,"children":353},{"className":352},[],[354],{"type":45,"value":267},{"type":40,"tag":229,"props":356,"children":357},{},[358],{"type":40,"tag":233,"props":359,"children":360},{},[361,363,373],{"type":45,"value":362},"the\nID-encodes-properties rule in\n",{"type":40,"tag":69,"props":364,"children":366},{"href":365},"TASKS.md#use",[367],{"type":40,"tag":60,"props":368,"children":370},{"className":369},[],[371],{"type":45,"value":372},"TASKS.md ## use",{"type":45,"value":374}," (a Data ID carries\nthe counter's property dimensions; the reverse-resolve\nreports them).",{"type":40,"tag":233,"props":376,"children":377},{},[378,383,384,420,422,430],{"type":40,"tag":52,"props":379,"children":380},{},[381],{"type":45,"value":382},"\"What property dimensions does this counter take and\nwhat values are valid?\"",{"type":45,"value":297},{"type":40,"tag":116,"props":385,"children":386},{},[387,389,395,397,403,405,411,412,418],{"type":45,"value":388},"\"I know\nthe counter is ",{"type":40,"tag":60,"props":390,"children":392},{"className":391},[],[393],{"type":45,"value":394},"pcie_link_write_stalled_time_*",{"type":45,"value":396}," — what\ndo ",{"type":40,"tag":60,"props":398,"children":400},{"className":399},[],[401],{"type":45,"value":402},"node",{"type":45,"value":404}," \u002F ",{"type":40,"tag":60,"props":406,"children":408},{"className":407},[],[409],{"type":45,"value":410},"pcie_index",{"type":45,"value":404},{"type":40,"tag":60,"props":413,"children":415},{"className":414},[],[416],{"type":45,"value":417},"depth",{"type":45,"value":419}," mean and what\nvalues does the device accept?\"",{"type":45,"value":421},". Answered by the\nproperty-dimension table in\n",{"type":40,"tag":69,"props":423,"children":424},{"href":260},[425],{"type":40,"tag":60,"props":426,"children":428},{"className":427},[],[429],{"type":45,"value":267},{"type":40,"tag":229,"props":431,"children":432},{},[433],{"type":40,"tag":233,"props":434,"children":435},{},[436,438,444],{"type":45,"value":437},"the ",{"type":40,"tag":60,"props":439,"children":441},{"className":440},[],[442],{"type":45,"value":443},"\u003Cname>",{"type":45,"value":445}," invocation without arguments which\nprints the documented property options + units +\nunit-specific axes.",{"type":40,"tag":233,"props":447,"children":448},{},[449,454,456,476,478,484,486,494],{"type":40,"tag":52,"props":450,"children":451},{},[452],{"type":45,"value":453},"\"Is this counter supported on this device before I\ncommit it to the exporter config?\"",{"type":45,"value":455}," — worked example:\n",{"type":40,"tag":116,"props":457,"children":458},{},[459,461,466,468,474],{"type":45,"value":460},"\"validate that ",{"type":40,"tag":60,"props":462,"children":464},{"className":463},[],[465],{"type":45,"value":246},{"type":45,"value":467}," with ",{"type":40,"tag":60,"props":469,"children":471},{"className":470},[],[472],{"type":45,"value":473},"node=1",{"type":45,"value":475}," is\nexposed on the BlueField at PCIe address X before I\nwrite the exporter config\"",{"type":45,"value":477},". Answered by the\nper-device-support probe (",{"type":40,"tag":60,"props":479,"children":481},{"className":480},[],[482],{"type":45,"value":483},"\u003Cdevice PCI> \u003Cname> [properties]",{"type":45,"value":485},") in\n",{"type":40,"tag":69,"props":487,"children":488},{"href":260},[489],{"type":40,"tag":60,"props":490,"children":492},{"className":491},[],[493],{"type":45,"value":267},{"type":40,"tag":229,"props":495,"children":496},{},[497],{"type":40,"tag":233,"props":498,"children":499},{},[500,502,510],{"type":45,"value":501},"the gate-before-commit rule in\n",{"type":40,"tag":69,"props":503,"children":504},{"href":365},[505],{"type":40,"tag":60,"props":506,"children":508},{"className":507},[],[509],{"type":45,"value":372},{"type":45,"value":154},{"type":40,"tag":233,"props":512,"children":513},{},[514,533,534,539,541,551,553,563,565,577],{"type":40,"tag":52,"props":515,"children":516},{},[517,519,524,526,531],{"type":45,"value":518},"\"Is ",{"type":40,"tag":60,"props":520,"children":522},{"className":521},[],[523],{"type":45,"value":65},{"type":45,"value":525}," on my install, and is it\npaired with the matching ",{"type":40,"tag":60,"props":527,"children":529},{"className":528},[],[530],{"type":45,"value":137},{"type":45,"value":532}," library\nversion?\"",{"type":45,"value":297},{"type":40,"tag":116,"props":535,"children":536},{},[537],{"type":45,"value":538},"\"is the diagnostic-\ndata counter set my exporter targets on this DOCA\nversion?\"",{"type":45,"value":540},". Answered by the version-overlay in\n",{"type":40,"tag":69,"props":542,"children":544},{"href":543},"CAPABILITIES.md#version-compatibility",[545],{"type":40,"tag":60,"props":546,"children":548},{"className":547},[],[549],{"type":45,"value":550},"CAPABILITIES.md ## Version compatibility",{"type":45,"value":552},",\nwhich redirects to the canonical\n",{"type":40,"tag":69,"props":554,"children":556},{"href":555},"..\u002F..\u002Fdoca-version\u002FSKILL.md",[557],{"type":40,"tag":60,"props":558,"children":560},{"className":559},[],[561],{"type":45,"value":562},"doca-version",{"type":45,"value":564}," chain\nand adds the ",{"type":40,"tag":116,"props":566,"children":567},{},[568,570,575],{"type":45,"value":569},"tool ↔ ",{"type":40,"tag":60,"props":571,"children":573},{"className":572},[],[574],{"type":45,"value":137},{"type":45,"value":576}," library\nschema-version",{"type":45,"value":578}," match rule.",{"type":40,"tag":210,"props":580,"children":582},{"id":581},"audience",[583],{"type":45,"value":584},"Audience",{"type":40,"tag":48,"props":586,"children":587},{},[588,590,595],{"type":45,"value":589},"This skill serves ",{"type":40,"tag":52,"props":591,"children":592},{},[593],{"type":45,"value":594},"external operators, developers, and\nAI agents standing up or debugging a DOCA Telemetry\nexporter \u002F collector pipeline who need to confirm the\ncounter schema, validate per-device support, or\ntranslate between human-readable counter names and the\nbinary Data IDs the exporter actually ships",{"type":45,"value":596},".\nConcretely:",{"type":40,"tag":229,"props":598,"children":599},{},[600,605,620,645],{"type":40,"tag":233,"props":601,"children":602},{},[603],{"type":45,"value":604},"A platform operator standing up a new DOCA Telemetry\nexporter on a BlueField fleet who needs to confirm\nwhich counters the target devices actually expose\nbefore committing the exporter config.",{"type":40,"tag":233,"props":606,"children":607},{},[608,610,618],{"type":45,"value":609},"A developer of a downstream consumer (a collector\napp linking ",{"type":40,"tag":69,"props":611,"children":612},{"href":130},[613],{"type":40,"tag":60,"props":614,"children":616},{"className":615},[],[617],{"type":45,"value":137},{"type":45,"value":619},",\nor a third-party aggregator consuming via DTS) who\nhas a captured Data ID stream and needs to translate\nIDs back to counter names + properties.",{"type":40,"tag":233,"props":621,"children":622},{},[623,625,630,631,636,638,643],{"type":45,"value":624},"An operator debugging a ",{"type":40,"tag":116,"props":626,"children":627},{},[628],{"type":45,"value":629},"\"nothing is shipping\ndownstream\"",{"type":45,"value":404},{"type":40,"tag":116,"props":632,"children":633},{},[634],{"type":45,"value":635},"\"this metric is silently missing\"",{"type":45,"value":637},"\nreport against a deployed exporter — the\nschema-discovery + per-device-support probes are the\ncanonical ",{"type":40,"tag":116,"props":639,"children":640},{},[641],{"type":45,"value":642},"\"is the counter even supposed to work on\nthis device\"",{"type":45,"value":644}," first step before suspecting the\ncollector or the network path.",{"type":40,"tag":233,"props":646,"children":647},{},[648,650,655],{"type":45,"value":649},"An AI agent producing a ",{"type":40,"tag":116,"props":651,"children":652},{},[653],{"type":45,"value":654},"\"validated exporter config\nfor this BlueField + this DOCA version\"",{"type":45,"value":656}," answer\nhonestly — with each counter resolved to its Data ID\nand each Data ID confirmed against the per-device\ncapability probe.",{"type":40,"tag":48,"props":658,"children":659},{},[660,662,667,669,674,676,680,682,686,688,693,695,706,708,716],{"type":45,"value":661},"It is ",{"type":40,"tag":52,"props":663,"children":664},{},[665],{"type":45,"value":666},"not",{"type":45,"value":668}," for users debugging the ",{"type":40,"tag":60,"props":670,"children":672},{"className":671},[],[673],{"type":45,"value":65},{"type":45,"value":675},"\nbinary itself, ",{"type":40,"tag":52,"props":677,"children":678},{},[679],{"type":45,"value":666},{"type":45,"value":681}," a substitute for the live public\nDOCA Telemetry guides, and ",{"type":40,"tag":52,"props":683,"children":684},{},[685],{"type":45,"value":666},{"type":45,"value":687}," the right place for\nlearning how to ",{"type":40,"tag":116,"props":689,"children":690},{},[691],{"type":45,"value":692},"write",{"type":45,"value":694}," an exporter application (that\naudience belongs in\n",{"type":40,"tag":69,"props":696,"children":698},{"href":697},"..\u002F..\u002Flibs\u002Fdoca-telemetry-exporter\u002FSKILL.md",[699,704],{"type":40,"tag":60,"props":700,"children":702},{"className":701},[],[703],{"type":45,"value":182},{"type":45,"value":705}," skill",{"type":45,"value":707},"\nwhen present, or via\n",{"type":40,"tag":69,"props":709,"children":710},{"href":199},[711],{"type":40,"tag":60,"props":712,"children":714},{"className":713},[],[715],{"type":45,"value":206},{"type":45,"value":717},").",{"type":40,"tag":48,"props":719,"children":720},{},[721,723,728,730,736,738,744],{"type":45,"value":722},"The tool is shipped as a ",{"type":40,"tag":52,"props":724,"children":725},{},[726],{"type":45,"value":727},"CLI binary",{"type":45,"value":729}," under\n",{"type":40,"tag":60,"props":731,"children":733},{"className":732},[],[734],{"type":45,"value":735},"\u002Fopt\u002Fmellanox\u002Fdoca\u002Ftools\u002F",{"type":45,"value":737},", not a library you link\nagainst. The skill uses the same ",{"type":40,"tag":60,"props":739,"children":741},{"className":740},[],[742],{"type":45,"value":743},"kind: tool",{"type":45,"value":745},"\nthree-file shape as the rest of the bundle so the\nagent's task-verb contract is uniform across libraries,\nservices, and tools.",{"type":40,"tag":210,"props":747,"children":749},{"id":748},"language-scope",[750],{"type":45,"value":751},"Language scope",{"type":40,"tag":48,"props":753,"children":754},{},[755,760,762,768],{"type":40,"tag":60,"props":756,"children":758},{"className":757},[],[759],{"type":45,"value":65},{"type":45,"value":761}," is a C host-side CLI that uses\nthe documented ",{"type":40,"tag":60,"props":763,"children":765},{"className":764},[],[766],{"type":45,"value":767},"doca_telemetry_diag",{"type":45,"value":769}," surface to list\ncounter types and probe per-device support. Its inputs\nare command-line arguments (counter name, property\nvalues, Data ID, optional device PCI address); its\noutputs are human-readable Data ID + properties\nmappings. The skill keeps workflow guidance language-\nneutral; downstream consumers in any language can use\nthe resolved Data IDs against their own collector code.",{"type":40,"tag":210,"props":771,"children":773},{"id":772},"when-to-load-this-skill",[774],{"type":45,"value":775},"When to load this skill",{"type":40,"tag":48,"props":777,"children":778},{},[779,781,786],{"type":45,"value":780},"Load this skill when the user is — or the agent needs\nto — invoke ",{"type":40,"tag":60,"props":782,"children":784},{"className":783},[],[785],{"type":45,"value":65},{"type":45,"value":787}," on a real host with\nDOCA installed, alongside an exporter \u002F collector\npipeline that needs schema discovery, per-device\nsupport validation, or Data ID translation.\nConcretely:",{"type":40,"tag":229,"props":789,"children":790},{},[791,803,808,813,825,837],{"type":40,"tag":233,"props":792,"children":793},{},[794,796,802],{"type":45,"value":795},"Enumerating the full counter schema before writing\na fresh DOCA Telemetry exporter config (",{"type":40,"tag":60,"props":797,"children":799},{"className":798},[],[800],{"type":45,"value":801},"get-counters",{"type":45,"value":717},{"type":40,"tag":233,"props":804,"children":805},{},[806],{"type":45,"value":807},"Validating that a chosen counter + property set\nresolves to a Data ID the target device actually\nsupports, before committing the config.",{"type":40,"tag":233,"props":809,"children":810},{},[811],{"type":45,"value":812},"Reverse-resolving a Data ID captured from a\ndownstream consumer back to a counter name +\nproperties for correlation against the public guide.",{"type":40,"tag":233,"props":814,"children":815},{},[816,818,823],{"type":45,"value":817},"Debugging a ",{"type":40,"tag":116,"props":819,"children":820},{},[821],{"type":45,"value":822},"\"exporter ships, collector receives\nnothing\"",{"type":45,"value":824}," report (the canonical schema-mismatch \u002F\nunsupported-counter failure mode).",{"type":40,"tag":233,"props":826,"children":827},{},[828,830,835],{"type":45,"value":829},"Producing a ",{"type":40,"tag":116,"props":831,"children":832},{},[833],{"type":45,"value":834},"\"validated counter set for this\nBlueField + this DOCA version\"",{"type":45,"value":836}," artifact as part of\na structured exporter-config baseline.",{"type":40,"tag":233,"props":838,"children":839},{},[840],{"type":45,"value":841},"Migrating an exporter pipeline across DOCA versions\nand confirming each previously-supported counter\nstill resolves cleanly on the new version.",{"type":40,"tag":48,"props":843,"children":844},{},[845,847,851,853,861,863,871,873,883],{"type":45,"value":846},"Do ",{"type":40,"tag":52,"props":848,"children":849},{},[850],{"type":45,"value":666},{"type":45,"value":852}," load this skill for general DOCA orientation,\ncollector \u002F exporter library programming, DTS\ndeployment, or DOCA install. For those, route to\n",{"type":40,"tag":69,"props":854,"children":855},{"href":199},[856],{"type":40,"tag":60,"props":857,"children":859},{"className":858},[],[860],{"type":45,"value":206},{"type":45,"value":862},",\n",{"type":40,"tag":69,"props":864,"children":865},{"href":130},[866],{"type":40,"tag":60,"props":867,"children":869},{"className":868},[],[870],{"type":45,"value":137},{"type":45,"value":872},",\nor ",{"type":40,"tag":69,"props":874,"children":876},{"href":875},"..\u002F..\u002Fdoca-setup\u002FSKILL.md",[877],{"type":40,"tag":60,"props":878,"children":880},{"className":879},[],[881],{"type":45,"value":882},"doca-setup",{"type":45,"value":154},{"type":40,"tag":210,"props":885,"children":887},{"id":886},"what-this-skill-provides",[888],{"type":45,"value":889},"What this skill provides",{"type":40,"tag":48,"props":891,"children":892},{},[893,895,900],{"type":45,"value":894},"This is a ",{"type":40,"tag":52,"props":896,"children":897},{},[898],{"type":45,"value":899},"thin loader",{"type":45,"value":901},". Substantive material lives\nin two companion files:",{"type":40,"tag":229,"props":903,"children":904},{},[905,956],{"type":40,"tag":233,"props":906,"children":907},{},[908,913,915,920,922,927,928,933,934,939,941,947,949,954],{"type":40,"tag":60,"props":909,"children":911},{"className":910},[],[912],{"type":45,"value":106},{"type":45,"value":914}," — what ",{"type":40,"tag":60,"props":916,"children":918},{"className":917},[],[919],{"type":45,"value":65},{"type":45,"value":921},"\ndiscovers + how it pairs with the developer-side\nsurfaces: the three documented invocation classes\n(enumerate \u002F name→ID \u002F ID→name), the property-\ndimension model (counters carry property axes such\nas ",{"type":40,"tag":60,"props":923,"children":925},{"className":924},[],[926],{"type":45,"value":402},{"type":45,"value":122},{"type":40,"tag":60,"props":929,"children":931},{"className":930},[],[932],{"type":45,"value":410},{"type":45,"value":122},{"type":40,"tag":60,"props":935,"children":937},{"className":936},[],[938],{"type":45,"value":417},{"type":45,"value":940},", plus per-unit\naxes), the optional per-device capability probe\n(",{"type":40,"tag":60,"props":942,"children":944},{"className":943},[],[945],{"type":45,"value":946},"\u003Cdevice PCI> \u003Cname>",{"type":45,"value":948}," runs the resolved counter\nagainst the device), the operator-side support\nrole (this is NOT the developer-side library; it\nexists to make exporter \u002F collector setups\nhonest), the version overlay (tool ↔\n",{"type":40,"tag":60,"props":950,"children":952},{"className":951},[],[953],{"type":45,"value":137},{"type":45,"value":955}," library schema version pairing),\nthe layered error taxonomy (install \u002F parse \u002F\nunknown-counter \u002F unknown-data-id \u002F property-out-\nof-range \u002F device-not-supported \u002F version \u002F\ncross-cutting), the observability surface\n(stdout-only), and the safety policy (the tool is\nread-only; mistakes appear downstream as silent\nmetric drops, not crashes).",{"type":40,"tag":233,"props":957,"children":958},{},[959,964,966,972,974,980,982,988,990,996,998,1004,1006,1012,1014,1020,1022,1028,1030,1035,1037,1043],{"type":40,"tag":60,"props":960,"children":962},{"className":961},[],[963],{"type":45,"value":71},{"type":45,"value":965}," — step-by-step workflows for the\nin-scope task verbs: ",{"type":40,"tag":60,"props":967,"children":969},{"className":968},[],[970],{"type":45,"value":971},"install",{"type":45,"value":973}," (host-side DOCA +\ntelemetry component prerequisites), ",{"type":40,"tag":60,"props":975,"children":977},{"className":976},[],[978],{"type":45,"value":979},"configure",{"type":45,"value":981},"\n(axis decisions: which invocation class, which\ndevice, which counter), ",{"type":40,"tag":60,"props":983,"children":985},{"className":984},[],[986],{"type":45,"value":987},"build",{"type":45,"value":989}," (route to install\n— the binary is shipped), ",{"type":40,"tag":60,"props":991,"children":993},{"className":992},[],[994],{"type":45,"value":995},"modify",{"type":45,"value":997}," (refuse — do\nnot patch the binary; modify the invocation and\nthe exporter \u002F collector config that consumes the\nresolved Data IDs), ",{"type":40,"tag":60,"props":999,"children":1001},{"className":1000},[],[1002],{"type":45,"value":1003},"run",{"type":45,"value":1005}," (the three documented\ninvocations), ",{"type":40,"tag":60,"props":1007,"children":1009},{"className":1008},[],[1010],{"type":45,"value":1011},"test",{"type":45,"value":1013}," (round-trip a chosen\ncounter through name → Data ID → per-device\nprobe → exporter config → collector receipt),\n",{"type":40,"tag":60,"props":1015,"children":1017},{"className":1016},[],[1018],{"type":45,"value":1019},"debug",{"type":45,"value":1021}," (walk the error taxonomy), ",{"type":40,"tag":60,"props":1023,"children":1025},{"className":1024},[],[1026],{"type":45,"value":1027},"use",{"type":45,"value":1029}," (the\nhand-off into the developer-side ",{"type":40,"tag":60,"props":1031,"children":1033},{"className":1032},[],[1034],{"type":45,"value":137},{"type":45,"value":1036},"\ncollector \u002F exporter pipeline), plus a ",{"type":40,"tag":60,"props":1038,"children":1040},{"className":1039},[],[1041],{"type":45,"value":1042},"Deferred task verbs",{"type":45,"value":1044}," block.",{"type":40,"tag":48,"props":1046,"children":1047},{},[1048,1050,1058],{"type":45,"value":1049},"The skill assumes a host where DOCA is already\ninstalled with the telemetry component, a BlueField\nthe operator is targeting is visible to DOCA, and the\nexporter \u002F collector pipeline that consumes the\nresolved Data IDs is either being authored or has\nalready been authored against the matching\n",{"type":40,"tag":69,"props":1051,"children":1052},{"href":130},[1053],{"type":40,"tag":60,"props":1054,"children":1056},{"className":1055},[],[1057],{"type":45,"value":137},{"type":45,"value":1059},"\nlibrary version.",{"type":40,"tag":210,"props":1061,"children":1063},{"id":1062},"what-this-skill-deliberately-does-not-ship",[1064],{"type":45,"value":1065},"What this skill deliberately does not ship",{"type":40,"tag":48,"props":1067,"children":1068},{},[1069,1071,1076],{"type":45,"value":1070},"This skill is ",{"type":40,"tag":52,"props":1072,"children":1073},{},[1074],{"type":45,"value":1075},"agent guidance",{"type":45,"value":1077},", not a samples or\nscripts bundle. To keep the boundary clean, it\ndeliberately does not contain — and pull requests\nshould not add:",{"type":40,"tag":229,"props":1079,"children":1080},{},[1081,1098,1108,1135,1145],{"type":40,"tag":233,"props":1082,"children":1083},{},[1084,1089,1091,1096],{"type":40,"tag":52,"props":1085,"children":1086},{},[1087],{"type":45,"value":1088},"A verbatim counter inventory or Data ID\ntable.",{"type":45,"value":1090}," The counter set evolves across DOCA\nversions and BlueField generations; an\ninventory pinned in this skill would silently rot.\n",{"type":40,"tag":60,"props":1092,"children":1094},{"className":1093},[],[1095],{"type":45,"value":320},{"type":45,"value":1097}," on the\ninstalled binary is the authoritative source.",{"type":40,"tag":233,"props":1099,"children":1100},{},[1101,1106],{"type":40,"tag":52,"props":1102,"children":1103},{},[1104],{"type":45,"value":1105},"Pre-baked exporter \u002F collector configs.",{"type":45,"value":1107}," Configs\nare install-, device-, and use-case-specific; a\npackaged config in this skill would mislead\noperators on a different setup.",{"type":40,"tag":233,"props":1109,"children":1110},{},[1111,1116,1118,1126,1128,1133],{"type":40,"tag":52,"props":1112,"children":1113},{},[1114],{"type":45,"value":1115},"A DTS deployment recipe.",{"type":45,"value":1117}," DTS is a separate DOCA\nservice with its own public guide; route via\n",{"type":40,"tag":69,"props":1119,"children":1120},{"href":199},[1121],{"type":40,"tag":60,"props":1122,"children":1124},{"className":1123},[],[1125],{"type":45,"value":206},{"type":45,"value":1127},".\nThis skill ",{"type":40,"tag":116,"props":1129,"children":1130},{},[1131],{"type":45,"value":1132},"resolves the counter names + IDs",{"type":45,"value":1134}," a\nDTS pipeline consumes; it does not configure DTS.",{"type":40,"tag":233,"props":1136,"children":1137},{},[1138,1143],{"type":40,"tag":52,"props":1139,"children":1140},{},[1141],{"type":45,"value":1142},"Wrappers, parsers, or scripts",{"type":45,"value":1144}," in any language\nthat consume the tool's output. The output is a\nsimple text mapping; users who want to script\nagainst it should read the live guide and write\nthe parser against their installed version.",{"type":40,"tag":233,"props":1146,"children":1147},{},[1148,1169,1171,1177,1179,1187],{"type":40,"tag":52,"props":1149,"children":1150},{},[1151,1153,1159,1161,1167],{"type":45,"value":1152},"A ",{"type":40,"tag":60,"props":1154,"children":1156},{"className":1155},[],[1157],{"type":45,"value":1158},"samples\u002F",{"type":45,"value":1160}," or ",{"type":40,"tag":60,"props":1162,"children":1164},{"className":1163},[],[1165],{"type":45,"value":1166},"reference\u002F",{"type":45,"value":1168}," subtree.",{"type":45,"value":1170}," This is\na thin loader for a shipped CLI; substantive\nmaterial lives on the public page, in ",{"type":40,"tag":60,"props":1172,"children":1174},{"className":1173},[],[1175],{"type":45,"value":1176},"--help",{"type":45,"value":1178},",\nand in ",{"type":40,"tag":69,"props":1180,"children":1181},{"href":130},[1182],{"type":40,"tag":60,"props":1183,"children":1185},{"className":1184},[],[1186],{"type":45,"value":137},{"type":45,"value":154},{"type":40,"tag":210,"props":1189,"children":1191},{"id":1190},"loading-order",[1192],{"type":45,"value":1193},"Loading order",{"type":40,"tag":1195,"props":1196,"children":1197},"ol",{},[1198,1211,1224],{"type":40,"tag":233,"props":1199,"children":1200},{},[1201,1203,1209],{"type":45,"value":1202},"Read this ",{"type":40,"tag":60,"props":1204,"children":1206},{"className":1205},[],[1207],{"type":45,"value":1208},"SKILL.md",{"type":45,"value":1210}," first to confirm the user's\nquestion is in scope (operator-side support\ntooling for a telemetry pipeline, not the\ndeveloper-side library programming).",{"type":40,"tag":233,"props":1212,"children":1213},{},[1214],{"type":40,"tag":52,"props":1215,"children":1216},{},[1217,1219,1223],{"type":45,"value":1218},"For the three invocation classes, the\nproperty-dimension model, the per-device support\nprobe, the version overlay, the error taxonomy,\nthe observability surface, and the safety\npolicy, see ",{"type":40,"tag":69,"props":1220,"children":1221},{"href":106},[1222],{"type":45,"value":106},{"type":45,"value":154},{"type":40,"tag":233,"props":1225,"children":1226},{},[1227],{"type":40,"tag":52,"props":1228,"children":1229},{},[1230,1232,1237,1238,1243,1244,1249,1250,1255,1256,1261,1262,1267,1268,1273,1274,1279,1280,1284],{"type":45,"value":1231},"For the documented invocations and the\ndiscover → resolve → validate → consume\nworkflow — ",{"type":40,"tag":60,"props":1233,"children":1235},{"className":1234},[],[1236],{"type":45,"value":971},{"type":45,"value":122},{"type":40,"tag":60,"props":1239,"children":1241},{"className":1240},[],[1242],{"type":45,"value":979},{"type":45,"value":122},{"type":40,"tag":60,"props":1245,"children":1247},{"className":1246},[],[1248],{"type":45,"value":987},{"type":45,"value":862},{"type":40,"tag":60,"props":1251,"children":1253},{"className":1252},[],[1254],{"type":45,"value":995},{"type":45,"value":122},{"type":40,"tag":60,"props":1257,"children":1259},{"className":1258},[],[1260],{"type":45,"value":1003},{"type":45,"value":122},{"type":40,"tag":60,"props":1263,"children":1265},{"className":1264},[],[1266],{"type":45,"value":1011},{"type":45,"value":122},{"type":40,"tag":60,"props":1269,"children":1271},{"className":1270},[],[1272],{"type":45,"value":1019},{"type":45,"value":122},{"type":40,"tag":60,"props":1275,"children":1277},{"className":1276},[],[1278],{"type":45,"value":1027},{"type":45,"value":184},{"type":40,"tag":69,"props":1281,"children":1282},{"href":71},[1283],{"type":45,"value":71},{"type":45,"value":154},{"type":40,"tag":210,"props":1286,"children":1288},{"id":1287},"related-skills",[1289],{"type":45,"value":1290},"Related skills",{"type":40,"tag":229,"props":1292,"children":1293},{},[1294,1314,1335,1379,1399,1414,1439],{"type":40,"tag":233,"props":1295,"children":1296},{},[1297,1305,1307,1312],{"type":40,"tag":69,"props":1298,"children":1299},{"href":130},[1300],{"type":40,"tag":60,"props":1301,"children":1303},{"className":1302},[],[1304],{"type":45,"value":137},{"type":45,"value":1306},"\n— the developer-side collector library whose\nschema this tool helps the operator discover. Pair\nthem in every exporter-pipeline triage session.\nThe collector library skill teaches the\ncollector-vs-exporter rule, the schema-must-match\ncontract with the publisher, and the consumer-\nqueue-full back-pressure rule; this tool's role\nis to make the ",{"type":40,"tag":116,"props":1308,"children":1309},{},[1310],{"type":45,"value":1311},"schema",{"type":45,"value":1313}," half of that contract\ninspectable from the operator side. Conflating\nthe library and the tool is the most common\ntelemetry first-touch error.",{"type":40,"tag":233,"props":1315,"children":1316},{},[1317,1325,1327,1333],{"type":40,"tag":69,"props":1318,"children":1319},{"href":199},[1320],{"type":40,"tag":60,"props":1321,"children":1323},{"className":1322},[],[1324],{"type":45,"value":206},{"type":45,"value":1326},"\n— routing to the public DOCA Telemetry guide and\nthe DOCA Telemetry Service (DTS) page on\n",{"type":40,"tag":60,"props":1328,"children":1330},{"className":1329},[],[1331],{"type":45,"value":1332},"docs.nvidia.com",{"type":45,"value":1334},", plus the on-disk install\nlayout for the tool.",{"type":40,"tag":233,"props":1336,"children":1337},{},[1338,1346,1348,1354,1356,1364,1366,1377],{"type":40,"tag":69,"props":1339,"children":1340},{"href":555},[1341],{"type":40,"tag":60,"props":1342,"children":1344},{"className":1343},[],[1345],{"type":45,"value":562},{"type":45,"value":1347}," —\ncanonical DOCA version-handling rules. The\n",{"type":40,"tag":60,"props":1349,"children":1351},{"className":1350},[],[1352],{"type":45,"value":1353},"## Version compatibility",{"type":45,"value":1355}," section in\n",{"type":40,"tag":69,"props":1357,"children":1358},{"href":106},[1359],{"type":40,"tag":60,"props":1360,"children":1362},{"className":1361},[],[1363],{"type":45,"value":106},{"type":45,"value":1365}," is a concise\noverlay that redirects here for the body and\nadds the ",{"type":40,"tag":116,"props":1367,"children":1368},{},[1369,1370,1375],{"type":45,"value":569},{"type":40,"tag":60,"props":1371,"children":1373},{"className":1372},[],[1374],{"type":45,"value":137},{"type":45,"value":1376}," library\nschema-version pairing",{"type":45,"value":1378}," rule.",{"type":40,"tag":233,"props":1380,"children":1381},{},[1382,1390,1392,1397],{"type":40,"tag":69,"props":1383,"children":1384},{"href":875},[1385],{"type":40,"tag":60,"props":1386,"children":1388},{"className":1387},[],[1389],{"type":45,"value":882},{"type":45,"value":1391}," — env\npreparation, install verification, and the ",{"type":40,"tag":116,"props":1393,"children":1394},{},[1395],{"type":45,"value":1396},"I\nhave no install yet",{"type":45,"value":1398}," path with the public NGC\nDOCA container. This skill assumes its\npreconditions are satisfied (DOCA installed\nwith the telemetry component, BlueField visible\nto DOCA).",{"type":40,"tag":233,"props":1400,"children":1401},{},[1402,1412],{"type":40,"tag":69,"props":1403,"children":1405},{"href":1404},"..\u002F..\u002Fdoca-debug\u002FSKILL.md",[1406],{"type":40,"tag":60,"props":1407,"children":1409},{"className":1408},[],[1410],{"type":45,"value":1411},"doca-debug",{"type":45,"value":1413}," — the\ncross-cutting debug ladder. Telemetry-utils\nfeeds the cross-cutting ladder by surfacing the\ncounter schema + per-device support truth at the\nruntime layer; exporter-pipeline regressions\noften resolve at the schema layer before\ntouching the collector \u002F network path.",{"type":40,"tag":233,"props":1415,"children":1416},{},[1417,1427,1429,1437],{"type":40,"tag":69,"props":1418,"children":1420},{"href":1419},"..\u002F..\u002Fdoca-structured-tools-contract\u002FSKILL.md",[1421],{"type":40,"tag":60,"props":1422,"children":1424},{"className":1423},[],[1425],{"type":45,"value":1426},"doca-structured-tools-contract",{"type":45,"value":1428},"\n— the bundle's detect → prefer → fall back →\nreport contract for structured helper tools. The\ncommand appendix in ",{"type":40,"tag":69,"props":1430,"children":1431},{"href":71},[1432],{"type":40,"tag":60,"props":1433,"children":1435},{"className":1434},[],[1436],{"type":45,"value":71},{"type":45,"value":1438},"\nhonors this contract.",{"type":40,"tag":233,"props":1440,"children":1441},{},[1442,1452,1454,1460],{"type":40,"tag":69,"props":1443,"children":1445},{"href":1444},"..\u002F..\u002Fdoca-programming-guide\u002FSKILL.md",[1446],{"type":40,"tag":60,"props":1447,"children":1449},{"className":1448},[],[1450],{"type":45,"value":1451},"doca-programming-guide",{"type":45,"value":1453},"\n— general DOCA programming patterns shared by\nevery library \u002F tool surface, including the\ncross-library ",{"type":40,"tag":60,"props":1455,"children":1457},{"className":1456},[],[1458],{"type":45,"value":1459},"DOCA_ERROR_*",{"type":45,"value":1461}," taxonomy this\ntool's host-side error layer overlays on top of\nwhen downstream collector \u002F exporter code\nsurfaces a related error.",{"items":1463,"total":1567},[1464,1481,1495,1509,1521,1538,1553],{"slug":1465,"name":1465,"fn":1466,"description":1467,"org":1468,"tags":1469,"stars":20,"repoUrl":21,"updatedAt":1480},"accelerated-computing-cudf","accelerate data processing with cuDF","Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV\u002FParquet I\u002FO, nullable semantics, and multi-GPU DataFrame workloads.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1470,1473,1476,1477],{"name":1471,"slug":1472,"type":15},"Data Analysis","data-analysis",{"name":1474,"slug":1475,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":1478,"slug":1479,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":1482,"name":1482,"fn":1483,"description":1484,"org":1485,"tags":1486,"stars":20,"repoUrl":21,"updatedAt":1494},"aiq-deploy","deploy and manage NVIDIA AI-Q infrastructure","Use when asked to install, deploy, run, validate, troubleshoot, or stop NVIDIA AI-Q Blueprint infrastructure.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1487,1490,1493],{"name":1488,"slug":1489,"type":15},"Deployment","deployment",{"name":1491,"slug":1492,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":1496,"name":1496,"fn":1497,"description":1498,"org":1499,"tags":1500,"stars":20,"repoUrl":21,"updatedAt":1508},"aiq-research","conduct deep research with AI-Q","Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1501,1504,1505],{"name":1502,"slug":1503,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":1506,"slug":1507,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":1510,"name":1510,"fn":1511,"description":1512,"org":1513,"tags":1514,"stars":20,"repoUrl":21,"updatedAt":1520},"amc-run-sample-calibration","run AMC sample dataset calibration","Run end-to-end calibration on the shipped sample dataset (sdg_08_2_sample_data_010926.zip) against a running AMC microservice. Use when user says 'test sample dataset', 'run sample calibration', 'verify AMC install', or 'launch and test'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1515,1516,1517],{"name":1471,"slug":1472,"type":15},{"name":9,"slug":8,"type":15},{"name":1518,"slug":1519,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":1522,"name":1522,"fn":1523,"description":1524,"org":1525,"tags":1526,"stars":20,"repoUrl":21,"updatedAt":1537},"amc-run-video-calibration","calibrate video datasets with AutoMagicCalib","Calibrate a new dataset from pre-recorded video files via the AutoMagicCalib REST API. Use when user has local MP4s and says 'calibrate my videos', 'run AMC on these videos', or similar. For RTSP\u002Flive streams, use amc-run-rtsp-calibration instead.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1527,1530,1533,1534],{"name":1528,"slug":1529,"type":15},"Automation","automation",{"name":1531,"slug":1532,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":1535,"slug":1536,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":1539,"name":1539,"fn":1540,"description":1541,"org":1542,"tags":1543,"stars":20,"repoUrl":21,"updatedAt":1552},"amc-setup-calibration-stack","deploy AutoMagicCalib microservice with Docker","Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1544,1545,1548,1549],{"name":1488,"slug":1489,"type":15},{"name":1546,"slug":1547,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":1550,"slug":1551,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":1554,"name":1554,"fn":1555,"description":1556,"org":1557,"tags":1558,"stars":20,"repoUrl":21,"updatedAt":1566},"cudaq-guide","develop quantum applications with CUDA-Q","CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1559,1560,1563],{"name":9,"slug":8,"type":15},{"name":1561,"slug":1562,"type":15},"Quantum Computing","quantum-computing",{"name":1564,"slug":1565,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":1569,"total":1720},[1570,1588,1604,1615,1627,1641,1654,1668,1679,1688,1702,1711],{"slug":1571,"name":1571,"fn":1572,"description":1573,"org":1574,"tags":1575,"stars":1585,"repoUrl":1586,"updatedAt":1587},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1576,1579,1582],{"name":1577,"slug":1578,"type":15},"Documentation","documentation",{"name":1580,"slug":1581,"type":15},"MCP","mcp",{"name":1583,"slug":1584,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1589,"name":1589,"fn":1590,"description":1591,"org":1592,"tags":1593,"stars":1601,"repoUrl":1602,"updatedAt":1603},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1594,1597,1598],{"name":1595,"slug":1596,"type":15},"Containers","containers",{"name":1488,"slug":1489,"type":15},{"name":1599,"slug":1600,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1605,"name":1605,"fn":1606,"description":1607,"org":1608,"tags":1609,"stars":1601,"repoUrl":1602,"updatedAt":1614},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1610,1613],{"name":1611,"slug":1612,"type":15},"CI\u002FCD","ci-cd",{"name":1488,"slug":1489,"type":15},"2026-07-14T05:25:59.97109",{"slug":1616,"name":1616,"fn":1617,"description":1618,"org":1619,"tags":1620,"stars":1601,"repoUrl":1602,"updatedAt":1626},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1621,1622,1623],{"name":1611,"slug":1612,"type":15},{"name":1488,"slug":1489,"type":15},{"name":1624,"slug":1625,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1628,"name":1628,"fn":1629,"description":1630,"org":1631,"tags":1632,"stars":1601,"repoUrl":1602,"updatedAt":1640},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1633,1636,1637],{"name":1634,"slug":1635,"type":15},"Debugging","debugging",{"name":1624,"slug":1625,"type":15},{"name":1638,"slug":1639,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1642,"name":1642,"fn":1643,"description":1644,"org":1645,"tags":1646,"stars":1601,"repoUrl":1602,"updatedAt":1653},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1647,1650],{"name":1648,"slug":1649,"type":15},"Best Practices","best-practices",{"name":1651,"slug":1652,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1655,"name":1655,"fn":1656,"description":1657,"org":1658,"tags":1659,"stars":1601,"repoUrl":1602,"updatedAt":1667},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1660,1663,1666],{"name":1661,"slug":1662,"type":15},"Machine Learning","machine-learning",{"name":1664,"slug":1665,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1669,"name":1669,"fn":1670,"description":1671,"org":1672,"tags":1673,"stars":1601,"repoUrl":1602,"updatedAt":1678},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1674,1677],{"name":1675,"slug":1676,"type":15},"QA","qa",{"name":1518,"slug":1519,"type":15},"2026-07-14T05:25:53.673039",{"slug":1680,"name":1680,"fn":1681,"description":1682,"org":1683,"tags":1684,"stars":1601,"repoUrl":1602,"updatedAt":1687},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1685,1686],{"name":1488,"slug":1489,"type":15},{"name":1491,"slug":1492,"type":15},"2026-07-14T05:25:49.362534",{"slug":1689,"name":1689,"fn":1690,"description":1691,"org":1692,"tags":1693,"stars":1601,"repoUrl":1602,"updatedAt":1701},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1694,1697,1698],{"name":1695,"slug":1696,"type":15},"Code Review","code-review",{"name":1624,"slug":1625,"type":15},{"name":1699,"slug":1700,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1703,"name":1703,"fn":1704,"description":1705,"org":1706,"tags":1707,"stars":1601,"repoUrl":1602,"updatedAt":1710},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1708,1709],{"name":1675,"slug":1676,"type":15},{"name":1518,"slug":1519,"type":15},"2026-07-14T05:25:54.928983",{"slug":1712,"name":1712,"fn":1713,"description":1714,"org":1715,"tags":1716,"stars":1601,"repoUrl":1602,"updatedAt":1719},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1717,1718],{"name":1528,"slug":1529,"type":15},{"name":1611,"slug":1612,"type":15},"2026-07-30T05:29:03.275638",496]