[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7AB5MPDV4TNC67QBCZPARNG":3,"skill-same:01M7AB5MPDV4TNC67QBCZPARNG":245},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":26,"commitSha":27,"skillDirectory":28,"forksCount":14,"repositoryHtmlUrl":29,"repositoryHomepageUrl":6,"ownerHtmlUrl":30,"ownerBlogUrl":6,"ownerDescription":31,"updatedAt":32,"install":33,"resources":46},"01M7AB5MPDV4TNC67QBCZPARNG","runpod",null,"Start here for any Runpod task — running GPU\u002FCPU pods, deploying serverless endpoints, templates, network volumes, building images, or understanding how Runpod works. Routes to the right skill (runpod-mcp, runpodctl, flash, companion-clis, runpod-usage, runpod-templates, runpod-migrate) and indexes two dozen live-verified end-to-end examples (golden paths) — use it when the lane is unclear, and for any multi-step or provisioning task even when it is not.","NeelakshSaxena\u002FVayu","NeelakshSaxena","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F87892838?v=4","01M7A3B9YFS6F6N9DD1787DT86","Neelaksh Saxena","01M7A3B9ZKBT09MNMPX2XN69BY",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: runpod\ndescription: >-\n  Start here for any Runpod task — running GPU\u002FCPU pods, deploying serverless\n  endpoints, templates, network volumes, building images, or understanding how\n  Runpod works. Routes to the right skill (runpod-mcp, runpodctl, flash,\n  companion-clis, runpod-usage, runpod-templates, runpod-migrate)\n  and indexes two dozen live-verified end-to-end examples\n  (golden paths) — use it when the lane is unclear, and for any multi-step or\n  provisioning task even when it is not.\nmetadata:\n  author: runpod\n  version: \"1.2.0\" # x-release-please-version\nlicense: Apache-2.0\n---\n\n# Runpod (router)\n\nThe entrypoint for the Runpod skills. This skill does no work itself — it picks\nthe right lane and hands off. For a multi-step or provisioning task, check the\n[worked examples](#worked-examples-golden-paths) first and let the matching one pick\nthe lane; otherwise read the matching skill's `SKILL.md` next.\n\n## The lanes\n\n| Lane | Use it for |\n| --- | --- |\n| **runpod-mcp** | Manage infra (pods, endpoints, jobs, templates, volumes, registries, catalog, billing) via **structured tool calls** — when the Runpod MCP tools are connected in this session. |\n| **runpodctl** | Manage the same infra from a **terminal\u002FCI\u002Fscript**, plus the things only the CLI does: Hub browse\u002Fdeploy, `send`\u002F`receive` file transfer, SSH keys, `doctor` setup, model cache. |\n| **flash** | **Write Python** that runs on Runpod serverless — `@remote`\u002F`@Endpoint` functions, `flash dev` hot-reload, `flash deploy`. Code-first, not infra management. |\n| **companion-clis** | **Prerequisite artifacts**: download a model (`hf`), build\u002Fpush an image (`docker`), repos\u002Freleases (`gh`), move data to a network volume over S3 (`aws`). |\n| **runpod-usage** | **Understand** how Runpod works before acting — pods vs serverless, building a container, storage, GPU selection, gotchas. Knowledge only. |\n| **runpod-templates** | **Official prebuilt pod templates** (ComfyUI, PyTorch, …): is there one for this workload, and what does its image ship — ports, paths, autostart, readiness, what's missing on first boot. Reference + routing hub; deploy via runpod-mcp\u002Frunpodctl. |\n| **runpod-migrate** | **Move existing code** off the GraphQL API or REST v1 onto REST v2 — inventory which parts use which version, rewrite call sites, flag breaking changes. Edits the user's code; does not manage infra. |\n\n## These skills are a snapshot; the tools are the source of truth\n\nEvery lane below wraps something that ships on its own release train and moves faster than\nthis repo. So route with these skills, but **take capability and syntax from the tool in front\nof you**:\n\n| Lane | Ask it directly |\n| --- | --- |\n| runpodctl | `runpodctl --help`, `runpodctl \u003Cresource> \u003Caction> --help`, `runpodctl version`. For failure shapes, run a command wrong on purpose and read the JSON |\n| runpod-mcp | your client's tool list (`\u002Fmcp` in Claude Code) — each tool carries its own parameter descriptions |\n| flash | `flash --help`, and the deploy\u002Fdev output |\n| companion-clis | that CLI's own `--help` (`hf`, `gh`, `docker`, `aws`) |\n| runpod-templates | `runpodctl template list --type official`, `template get \u003Cid>` (its readme is authoritative) |\n| REST v2 | the live spec at `https:\u002F\u002Fapi.runpod.io\u002Fv2\u002Fopenapi.json` |\n\n**Never tell a user a tool cannot do something without checking first.** A missing capability\nis the claim most likely to be out of date here, and it is the one a reader has no reason to\nreverify — it has already gone stale twice (runpodctl v2.9.0 added `serverless health`, v2.10.0\nadded `pod logs`\u002F`serverless logs`). If a limit still holds, name the version it holds for\nrather than saying \"cannot\".\n\n## First run — check auth before the first infra action\n\nInfra tasks (pods, endpoints, jobs, volumes) need a working control plane — the **Runpod MCP**\nor **runpodctl**. Don't start and discover mid-task that nothing's set up: check first, and if\nit isn't, help the user set up rather than limping on a partial fallback.\n\n**Check** (credential resolution order: `RUNPOD_API_KEY` env → `.env` → `~\u002F.runpod\u002Fconfig.toml`):\n```bash\nrunpodctl user            # succeeds ⇒ a key is set and valid\n```\nPlus, in Claude Code, `\u002Fmcp` should show `runpod` **Connected**.\n\n**Rule: get a key first — do not default to MCP OAuth.** The reason: one `RUNPOD_API_KEY`\nunlocks every tool — it authenticates **runpodctl + flash + the hosted MCP** (as `--header\n\"Authorization: Bearer $RUNPOD_API_KEY\"`). The MCP's \"Sign in with Runpod\" OAuth auths the **MCP alone** — the CLIs\nstay blocked, so you hit a wall on any CLI-only task (Hub, `send`\u002F`receive`, SSH, `doctor`,\nmodel cache\u002FModel Repository, CPU endpoints). ⚠️ **OAuth-only is a half-setup.** If nothing's\nset up, stop and get a key, in order:\n1. **`flash login`** — browser OAuth that saves a real key to `~\u002F.runpod\u002Fconfig.toml` (runpodctl\n   + flash read it; reuse it for the MCP Bearer). One step, unlocks all. Human-only.\n2. **`export RUNPOD_API_KEY=…`** (https:\u002F\u002Fconsole.runpod.io\u002Fuser\u002Fsettings) — same full unlock;\n   best for headless agents.\n3. **MCP OAuth only** (`\u002Fmcp` → *Sign in*) — last resort, MCP-only work; CLIs stay unauthed.\n\n**Then:** if a lane already works, use it — but if *only* the MCP is OAuth'd, still get a key\nbefore any CLI-only step. Missing a CLI? `curl -sSL https:\u002F\u002Fcli.runpod.net | bash` (runpodctl) ·\n`uv tool install runpod-flash` (flash) · `npx @runpod\u002Fmcp-server@latest add` (MCP). Full setup:\n[`runpod-usage\u002Freference\u002Fgetting-started.md`](..\u002Frunpod-usage\u002Freference\u002Fgetting-started.md).\n\n## How to route\n\n**0. Does a worked example already cover this?** For anything beyond a single call,\ncheck [the golden-paths index](#worked-examples-golden-paths) *before* picking a lane —\na matching path already names the lane(s), the flags, the ordering, and the traps, so\nrouting becomes reading rather than re-deriving. Check it when **any** of these is true:\n\n- the task needs **more than one resource** (image + template + endpoint, pod + volume,\n  multi-region, …) or **more than one lane**\n- it **provisions something billable**, or the user's ask is shaped like *\"get X running \u002F\n  deployed \u002F working\"*\n- it involves **storage, networking, autoscaling, streaming, or debugging a live resource** —\n  the areas where the non-obvious ordering is the whole difficulty\n- you are about to write a **multi-step plan** for it\n\nSkip step 0 for a single read or a single CRUD call (\"list my pods\", \"stop pod X\",\n\"what GPUs are available\") — go straight to the lane. **A matching golden path outranks\nthis router's lane table**: it was verified end to end on a real account, so where the two\ndisagree, follow the path and treat the difference as a bug worth reporting.\n\n1. **Conceptual question, or an unmade design choice** (serverless vs pod? which\n   GPU? bake the model or mount a volume?) → read **runpod-usage** first, then\n   continue with the answer.\n2. **Run a common workload on a pod, or fix a template pod** (\"run ComfyUI \u002F\n   PyTorch dev box\", won't boot, missing models) → **runpod-templates** — check for\n   an official prebuilt before planning any install, and let it route repairs.\n3. **Write\u002Fiterate\u002Fship your own code on Runpod GPUs** → **flash**.\n4. **Produce an artifact** (download a model, build+push an image, create a repo\n   release, sync data to a volume) → **companion-clis**.\n5. **Migrate existing code between Runpod API versions** — \"move us to REST v2\",\n   \"which Runpod API is this repo on?\", \"what breaks if we upgrade?\" →\n   **runpod-migrate**. (Calling the API to *do* something is a different job; that\n   is the infra lanes below.)\n6. **Manage infrastructure** (create\u002Flist\u002Fupdate\u002Fdelete pods, endpoints,\n   templates, volumes; list GPUs\u002Fdata centers; run a serverless job; billing):\n   - Capability only the CLI has — **Hub, `send`\u002F`receive`, SSH keys, `doctor`,\n     model cache** → **runpodctl**.\n   - Otherwise, if the Runpod **MCP tools are connected** in this session\n     (`create-pod`, `list-endpoints`, … are available) → **runpod-mcp**.\n   - Otherwise (shell-only agent, no MCP) → **runpodctl**.\n\n### runpod-mcp vs runpodctl (the overlap)\n\nBoth drive the same Runpod API, so they overlap on infra CRUD. Choose by\n**capability first, environment second**:\n\n- **MCP wins on convenience** for simple, structured operations — reads and basic\n  CRUD — when its tools are connected (typed params, no shell quoting).\n- **runpodctl takes over when an operation needs a capability MCP lacks** — even\n  if MCP is connected — and is the only option for a shell-only agent or when the\n  user wants a reproducible command.\n\nCapability matrix (pick the preferred lane per operation):\n\n| Operation | Preferred lane | Why |\n| --- | --- | --- |\n| List\u002Fget anything; start\u002Fstop\u002Frestart\u002Fdelete a pod; simple CRUD on endpoints, templates, volumes, registries; catalog; billing | **runpod-mcp** if connected, else runpodctl | Simple structured ops — MCP is typed and convenient |\n| Create a **simple** pod (one image + one GPU) | **runpod-mcp** if connected, else runpodctl | Both handle it |\n| Create a pod **from a template** or a **CPU** pod | **runpod-mcp** if connected, else runpodctl | MCP's create-pod takes `templateId` (v2-only) and `computeType: \"CPU\"` |\n| Create a pod with a **multi-GPU priority list**, or **template + CPU together** | **runpodctl** | MCP narrows to one GPU type, and rejects a template deploy for a CPU pod |\n| Deploy from the **Hub** | **runpod-mcp** if connected, else runpodctl | MCP has `list-hub-repos` + `deploy-hub-repo` |\n| **File transfer** (`send`\u002F`receive`), **SSH** keys\u002Finfo, **`doctor`** setup, **model** cache | **runpodctl** | MCP has no tool for these |\n| Invoke a serverless job (`run`\u002F`runsync`\u002Fstatus\u002Fstream) | **runpod-mcp** if connected, else runpodctl | Both lanes have first-class job commands now (runpodctl `serverless run`\u002F`status`\u002F`health`, v2.9.0+); MCP is typed, and only MCP streams a job's incremental output (`stream-job`) |\n| Read **pod or worker logs** | **either** — runpod-mcp if connected, else runpodctl | Both lanes read them: MCP `stream-pod-logs`\u002F`stream-worker-logs` return parsed frames; runpodctl `pod logs`\u002F`serverless logs` emit json lines and `--follow` (v2.10.0+) |\n\nRule of thumb: **default to MCP for the easy stuff, hand off to runpodctl the\nmoment an op needs a flag\u002Ffeature MCP doesn't expose.**\n\n## Deploying a workload (the golden loop)\n\nFor any \"get \u003CX> running on Runpod\" task, follow the **development loop** in\n`runpod-usage\u002Freference\u002Fdevelopment-loop.md`: decide pod vs serverless → provision → set up\n(only if from-scratch) → verify → deliver → cost-guard + teardown. Two rules bind within it:\n\n- **Prefer a prebuilt template \u002F Hub worker over building an image from scratch** —\n  official pod templates are indexed in [`runpod-templates`](..\u002Frunpod-templates\u002FSKILL.md).\n- **Before delivering, verify the workload with a real request from outside the pod\u002Fendpoint\n  — a \"Running\"\u002F\"ready\" status does not mean it is serving.**\n\nIt branches to two sub-loops:\n\n- **Service you open at a URL** (Ollama, ComfyUI, dev box) →\n  [`runpod-usage\u002Freference\u002Fpod-workflows.md`](..\u002Frunpod-usage\u002Freference\u002Fpod-workflows.md)\n  (ports + env + volume at creation, SSH-exec install, bind `0.0.0.0`, poll the\n  proxy URL). Execute in the runpodctl lane.\n- **Request\u002Fresponse API that scales to zero** (Whisper, inference) →\n  [`runpod-usage\u002Freference\u002Fendpoint-workflows.md`](..\u002Frunpod-usage\u002Freference\u002Fendpoint-workflows.md)\n  (Hub worker vs flash vs custom image; invoke `\u002Frun`\u002F`\u002Frunsync`; poll job status).\n\n## Worked examples (golden paths)\n\nThis is **step 0 of routing**, not an appendix. Two dozen end-to-end scenarios\n(nearly all **live-verified on a real account**) live in\n[`.\u002Fgolden-paths\u002FREADME.md`](.\u002Fgolden-paths\u002FREADME.md), with the real commands and the\nobserved output to copy from — plus a Gotchas and a Cost & cleanup section each, which\nis the part that is expensive to rediscover.\n\n**Match the task to a row below and open that path before you plan or call anything.**\nA partial match is still worth opening: the closest path's ordering and gotchas usually\ntransfer even when the model, GPU, or region does not. Only fall through to the lane\ntables above when nothing here is close.\n\n| Want to… | Golden path |\n| --- | --- |\n| Run a server (Ollama\u002FComfyUI) on a pod at a URL | [01](.\u002Fgolden-paths\u002F01-ollama-pod.md), [02](.\u002Fgolden-paths\u002F02-comfyui-pod\u002FREADME.md) |\n| Deploy a serverless model endpoint (Hub \u002F flash \u002F custom image) | [03](.\u002Fgolden-paths\u002F03-whisper-endpoint\u002FREADME.md), [05](.\u002Fgolden-paths\u002F05-model-to-endpoint-pipeline.md) |\n| Serve a HuggingFace model without baking it in or a volume (host-cached) | [20 — model caching (`--model-reference`)](.\u002Fgolden-paths\u002F20-model-caching-endpoint.md) |\n| Call a ready hosted model (no deploy) | [11 — Public Endpoints](.\u002Fgolden-paths\u002F11-public-endpoints.md) |\n| Fine-tune, then serve the result | [04](.\u002Fgolden-paths\u002F04-finetune-pod.md), [08](.\u002Fgolden-paths\u002F08-finetune-to-serverless.md) |\n| Interactive dev box (SSH \u002F VS Code) | [06](.\u002Fgolden-paths\u002F06-dev-pod.md) |\n| Move data pod → volume → serverless | [07](.\u002Fgolden-paths\u002F07-network-volume-handoff.md) |\n| **Custom serverless when flash isn't enough** (dual-mode image dev loop) | [09](.\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002FREADME.md) |\n| Build a minimal image for a target (pod vs serverless queue) | [22 (pod)](.\u002Fgolden-paths\u002F22-minimal-pod-image\u002FREADME.md), [23 (queue)](.\u002Fgolden-paths\u002F23-minimal-queue-image\u002FREADME.md); concepts in [building-images](..\u002Frunpod-usage\u002Freference\u002Fbuilding-images.md) |\n| Decide what to bake into the image vs mount on a network volume | [25 — bake vs mount](.\u002Fgolden-paths\u002F25-bake-vs-mount\u002FREADME.md) |\n| Pick a network-volume **storage tier** (standard vs high-performance) | [21 — storage tiers](.\u002Fgolden-paths\u002F21-storage-tiers.md) |\n| **High availability \u002F multi-region** serverless (multi-volume + data sync) | [10](.\u002Fgolden-paths\u002F10-multi-region-ha-serverless.md), [19 (3-region)](.\u002Fgolden-paths\u002F19-three-region-same-file.md) |\n| Stream output incrementally (`\u002Fstream`) | [12](.\u002Fgolden-paths\u002F12-serverless-streaming.md) |\n| Tune autoscaling \u002F raise per-worker throughput | [13 (autoscaling)](.\u002Fgolden-paths\u002F13-autoscaling-tuning.md), [18 (concurrency)](.\u002Fgolden-paths\u002F18-concurrent-handler.md) |\n| Load-balancing \u002F HTTP-server or WebSocket worker | [14 (LB)](.\u002Fgolden-paths\u002F14-load-balancing-endpoint.md), [17 (WebSocket)](.\u002Fgolden-paths\u002F17-serverless-websocket.md) |\n| Get notified on job completion (push, not poll) | [16 — webhooks](.\u002Fgolden-paths\u002F16-serverless-webhooks.md) |\n| Check health \u002F debug a failing endpoint | [15 — monitor & debug](.\u002Fgolden-paths\u002F15-monitor-and-debug.md) |\n\n## Multi-lane tasks\n\nSequence is always **understand → produce artifacts → manage infra → verify**,\nbecause infra can only reference artifacts that already exist. Keep each step in\none lane, and switch lanes at credential boundaries.\n\nExample — \"deploy `openai\u002Fgpt-oss-20b` to a serverless endpoint\":\n1. **runpod-usage** — serverless vs pod, GPU tier for 20B, bake vs mount vs cache.\n2. **companion-clis** — `hf download …`, `docker build --platform=linux\u002Famd64 …`, `docker push`.\n3. **runpod-mcp** or **runpodctl** — create the endpoint referencing the image + GPU pool.\n4. Same infra lane — invoke the endpoint \u002F check status to verify.\n\n## Auth\n\nEverything is one key: **`RUNPOD_API_KEY`** (https:\u002F\u002Fconsole.runpod.io\u002Fuser\u002Fsettings).\nEach lane just makes that key resolvable — `runpodctl doctor`, `flash login`, MCP\nstdio env var, or MCP hosted \"Sign in with Runpod\" (OAuth, no key on disk).\nCompanion CLIs use their **own** credentials (HuggingFace token, GitHub auth,\nDocker Hub PAT, Runpod **S3** keys for `aws`) — do not reuse `RUNPOD_API_KEY` for\nthose.\n","# Runpod (router)\n\nThe entrypoint for the Runpod skills. This skill does no work itself — it picks\nthe right lane and hands off. For a multi-step or provisioning task, check the\n[worked examples](#worked-examples-golden-paths) first and let the matching one pick\nthe lane; otherwise read the matching skill's `SKILL.md` next.\n\n## The lanes\n\n| Lane | Use it for |\n| --- | --- |\n| **runpod-mcp** | Manage infra (pods, endpoints, jobs, templates, volumes, registries, catalog, billing) via **structured tool calls** — when the Runpod MCP tools are connected in this session. |\n| **runpodctl** | Manage the same infra from a **terminal\u002FCI\u002Fscript**, plus the things only the CLI does: Hub browse\u002Fdeploy, `send`\u002F`receive` file transfer, SSH keys, `doctor` setup, model cache. |\n| **flash** | **Write Python** that runs on Runpod serverless — `@remote`\u002F`@Endpoint` functions, `flash dev` hot-reload, `flash deploy`. Code-first, not infra management. |\n| **companion-clis** | **Prerequisite artifacts**: download a model (`hf`), build\u002Fpush an image (`docker`), repos\u002Freleases (`gh`), move data to a network volume over S3 (`aws`). |\n| **runpod-usage** | **Understand** how Runpod works before acting — pods vs serverless, building a container, storage, GPU selection, gotchas. Knowledge only. |\n| **runpod-templates** | **Official prebuilt pod templates** (ComfyUI, PyTorch, …): is there one for this workload, and what does its image ship — ports, paths, autostart, readiness, what's missing on first boot. Reference + routing hub; deploy via runpod-mcp\u002Frunpodctl. |\n| **runpod-migrate** | **Move existing code** off the GraphQL API or REST v1 onto REST v2 — inventory which parts use which version, rewrite call sites, flag breaking changes. Edits the user's code; does not manage infra. |\n\n## These skills are a snapshot; the tools are the source of truth\n\nEvery lane below wraps something that ships on its own release train and moves faster than\nthis repo. So route with these skills, but **take capability and syntax from the tool in front\nof you**:\n\n| Lane | Ask it directly |\n| --- | --- |\n| runpodctl | `runpodctl --help`, `runpodctl \u003Cresource> \u003Caction> --help`, `runpodctl version`. For failure shapes, run a command wrong on purpose and read the JSON |\n| runpod-mcp | your client's tool list (`\u002Fmcp` in Claude Code) — each tool carries its own parameter descriptions |\n| flash | `flash --help`, and the deploy\u002Fdev output |\n| companion-clis | that CLI's own `--help` (`hf`, `gh`, `docker`, `aws`) |\n| runpod-templates | `runpodctl template list --type official`, `template get \u003Cid>` (its readme is authoritative) |\n| REST v2 | the live spec at `https:\u002F\u002Fapi.runpod.io\u002Fv2\u002Fopenapi.json` |\n\n**Never tell a user a tool cannot do something without checking first.** A missing capability\nis the claim most likely to be out of date here, and it is the one a reader has no reason to\nreverify — it has already gone stale twice (runpodctl v2.9.0 added `serverless health`, v2.10.0\nadded `pod logs`\u002F`serverless logs`). If a limit still holds, name the version it holds for\nrather than saying \"cannot\".\n\n## First run — check auth before the first infra action\n\nInfra tasks (pods, endpoints, jobs, volumes) need a working control plane — the **Runpod MCP**\nor **runpodctl**. Don't start and discover mid-task that nothing's set up: check first, and if\nit isn't, help the user set up rather than limping on a partial fallback.\n\n**Check** (credential resolution order: `RUNPOD_API_KEY` env → `.env` → `~\u002F.runpod\u002Fconfig.toml`):\n```bash\nrunpodctl user            # succeeds ⇒ a key is set and valid\n```\nPlus, in Claude Code, `\u002Fmcp` should show `runpod` **Connected**.\n\n**Rule: get a key first — do not default to MCP OAuth.** The reason: one `RUNPOD_API_KEY`\nunlocks every tool — it authenticates **runpodctl + flash + the hosted MCP** (as `--header\n\"Authorization: Bearer $RUNPOD_API_KEY\"`). The MCP's \"Sign in with Runpod\" OAuth auths the **MCP alone** — the CLIs\nstay blocked, so you hit a wall on any CLI-only task (Hub, `send`\u002F`receive`, SSH, `doctor`,\nmodel cache\u002FModel Repository, CPU endpoints). ⚠️ **OAuth-only is a half-setup.** If nothing's\nset up, stop and get a key, in order:\n1. **`flash login`** — browser OAuth that saves a real key to `~\u002F.runpod\u002Fconfig.toml` (runpodctl\n   + flash read it; reuse it for the MCP Bearer). One step, unlocks all. Human-only.\n2. **`export RUNPOD_API_KEY=…`** (https:\u002F\u002Fconsole.runpod.io\u002Fuser\u002Fsettings) — same full unlock;\n   best for headless agents.\n3. **MCP OAuth only** (`\u002Fmcp` → *Sign in*) — last resort, MCP-only work; CLIs stay unauthed.\n\n**Then:** if a lane already works, use it — but if *only* the MCP is OAuth'd, still get a key\nbefore any CLI-only step. Missing a CLI? `curl -sSL https:\u002F\u002Fcli.runpod.net | bash` (runpodctl) ·\n`uv tool install runpod-flash` (flash) · `npx @runpod\u002Fmcp-server@latest add` (MCP). Full setup:\n[`runpod-usage\u002Freference\u002Fgetting-started.md`](..\u002Frunpod-usage\u002Freference\u002Fgetting-started.md).\n\n## How to route\n\n**0. Does a worked example already cover this?** For anything beyond a single call,\ncheck [the golden-paths index](#worked-examples-golden-paths) *before* picking a lane —\na matching path already names the lane(s), the flags, the ordering, and the traps, so\nrouting becomes reading rather than re-deriving. Check it when **any** of these is true:\n\n- the task needs **more than one resource** (image + template + endpoint, pod + volume,\n  multi-region, …) or **more than one lane**\n- it **provisions something billable**, or the user's ask is shaped like *\"get X running \u002F\n  deployed \u002F working\"*\n- it involves **storage, networking, autoscaling, streaming, or debugging a live resource** —\n  the areas where the non-obvious ordering is the whole difficulty\n- you are about to write a **multi-step plan** for it\n\nSkip step 0 for a single read or a single CRUD call (\"list my pods\", \"stop pod X\",\n\"what GPUs are available\") — go straight to the lane. **A matching golden path outranks\nthis router's lane table**: it was verified end to end on a real account, so where the two\ndisagree, follow the path and treat the difference as a bug worth reporting.\n\n1. **Conceptual question, or an unmade design choice** (serverless vs pod? which\n   GPU? bake the model or mount a volume?) → read **runpod-usage** first, then\n   continue with the answer.\n2. **Run a common workload on a pod, or fix a template pod** (\"run ComfyUI \u002F\n   PyTorch dev box\", won't boot, missing models) → **runpod-templates** — check for\n   an official prebuilt before planning any install, and let it route repairs.\n3. **Write\u002Fiterate\u002Fship your own code on Runpod GPUs** → **flash**.\n4. **Produce an artifact** (download a model, build+push an image, create a repo\n   release, sync data to a volume) → **companion-clis**.\n5. **Migrate existing code between Runpod API versions** — \"move us to REST v2\",\n   \"which Runpod API is this repo on?\", \"what breaks if we upgrade?\" →\n   **runpod-migrate**. (Calling the API to *do* something is a different job; that\n   is the infra lanes below.)\n6. **Manage infrastructure** (create\u002Flist\u002Fupdate\u002Fdelete pods, endpoints,\n   templates, volumes; list GPUs\u002Fdata centers; run a serverless job; billing):\n   - Capability only the CLI has — **Hub, `send`\u002F`receive`, SSH keys, `doctor`,\n     model cache** → **runpodctl**.\n   - Otherwise, if the Runpod **MCP tools are connected** in this session\n     (`create-pod`, `list-endpoints`, … are available) → **runpod-mcp**.\n   - Otherwise (shell-only agent, no MCP) → **runpodctl**.\n\n### runpod-mcp vs runpodctl (the overlap)\n\nBoth drive the same Runpod API, so they overlap on infra CRUD. Choose by\n**capability first, environment second**:\n\n- **MCP wins on convenience** for simple, structured operations — reads and basic\n  CRUD — when its tools are connected (typed params, no shell quoting).\n- **runpodctl takes over when an operation needs a capability MCP lacks** — even\n  if MCP is connected — and is the only option for a shell-only agent or when the\n  user wants a reproducible command.\n\nCapability matrix (pick the preferred lane per operation):\n\n| Operation | Preferred lane | Why |\n| --- | --- | --- |\n| List\u002Fget anything; start\u002Fstop\u002Frestart\u002Fdelete a pod; simple CRUD on endpoints, templates, volumes, registries; catalog; billing | **runpod-mcp** if connected, else runpodctl | Simple structured ops — MCP is typed and convenient |\n| Create a **simple** pod (one image + one GPU) | **runpod-mcp** if connected, else runpodctl | Both handle it |\n| Create a pod **from a template** or a **CPU** pod | **runpod-mcp** if connected, else runpodctl | MCP's create-pod takes `templateId` (v2-only) and `computeType: \"CPU\"` |\n| Create a pod with a **multi-GPU priority list**, or **template + CPU together** | **runpodctl** | MCP narrows to one GPU type, and rejects a template deploy for a CPU pod |\n| Deploy from the **Hub** | **runpod-mcp** if connected, else runpodctl | MCP has `list-hub-repos` + `deploy-hub-repo` |\n| **File transfer** (`send`\u002F`receive`), **SSH** keys\u002Finfo, **`doctor`** setup, **model** cache | **runpodctl** | MCP has no tool for these |\n| Invoke a serverless job (`run`\u002F`runsync`\u002Fstatus\u002Fstream) | **runpod-mcp** if connected, else runpodctl | Both lanes have first-class job commands now (runpodctl `serverless run`\u002F`status`\u002F`health`, v2.9.0+); MCP is typed, and only MCP streams a job's incremental output (`stream-job`) |\n| Read **pod or worker logs** | **either** — runpod-mcp if connected, else runpodctl | Both lanes read them: MCP `stream-pod-logs`\u002F`stream-worker-logs` return parsed frames; runpodctl `pod logs`\u002F`serverless logs` emit json lines and `--follow` (v2.10.0+) |\n\nRule of thumb: **default to MCP for the easy stuff, hand off to runpodctl the\nmoment an op needs a flag\u002Ffeature MCP doesn't expose.**\n\n## Deploying a workload (the golden loop)\n\nFor any \"get \u003CX> running on Runpod\" task, follow the **development loop** in\n`runpod-usage\u002Freference\u002Fdevelopment-loop.md`: decide pod vs serverless → provision → set up\n(only if from-scratch) → verify → deliver → cost-guard + teardown. Two rules bind within it:\n\n- **Prefer a prebuilt template \u002F Hub worker over building an image from scratch** —\n  official pod templates are indexed in [`runpod-templates`](..\u002Frunpod-templates\u002FSKILL.md).\n- **Before delivering, verify the workload with a real request from outside the pod\u002Fendpoint\n  — a \"Running\"\u002F\"ready\" status does not mean it is serving.**\n\nIt branches to two sub-loops:\n\n- **Service you open at a URL** (Ollama, ComfyUI, dev box) →\n  [`runpod-usage\u002Freference\u002Fpod-workflows.md`](..\u002Frunpod-usage\u002Freference\u002Fpod-workflows.md)\n  (ports + env + volume at creation, SSH-exec install, bind `0.0.0.0`, poll the\n  proxy URL). Execute in the runpodctl lane.\n- **Request\u002Fresponse API that scales to zero** (Whisper, inference) →\n  [`runpod-usage\u002Freference\u002Fendpoint-workflows.md`](..\u002Frunpod-usage\u002Freference\u002Fendpoint-workflows.md)\n  (Hub worker vs flash vs custom image; invoke `\u002Frun`\u002F`\u002Frunsync`; poll job status).\n\n## Worked examples (golden paths)\n\nThis is **step 0 of routing**, not an appendix. Two dozen end-to-end scenarios\n(nearly all **live-verified on a real account**) live in\n[`.\u002Fgolden-paths\u002FREADME.md`](.\u002Fgolden-paths\u002FREADME.md), with the real commands and the\nobserved output to copy from — plus a Gotchas and a Cost & cleanup section each, which\nis the part that is expensive to rediscover.\n\n**Match the task to a row below and open that path before you plan or call anything.**\nA partial match is still worth opening: the closest path's ordering and gotchas usually\ntransfer even when the model, GPU, or region does not. Only fall through to the lane\ntables above when nothing here is close.\n\n| Want to… | Golden path |\n| --- | --- |\n| Run a server (Ollama\u002FComfyUI) on a pod at a URL | [01](.\u002Fgolden-paths\u002F01-ollama-pod.md), [02](.\u002Fgolden-paths\u002F02-comfyui-pod\u002FREADME.md) |\n| Deploy a serverless model endpoint (Hub \u002F flash \u002F custom image) | [03](.\u002Fgolden-paths\u002F03-whisper-endpoint\u002FREADME.md), [05](.\u002Fgolden-paths\u002F05-model-to-endpoint-pipeline.md) |\n| Serve a HuggingFace model without baking it in or a volume (host-cached) | [20 — model caching (`--model-reference`)](.\u002Fgolden-paths\u002F20-model-caching-endpoint.md) |\n| Call a ready hosted model (no deploy) | [11 — Public Endpoints](.\u002Fgolden-paths\u002F11-public-endpoints.md) |\n| Fine-tune, then serve the result | [04](.\u002Fgolden-paths\u002F04-finetune-pod.md), [08](.\u002Fgolden-paths\u002F08-finetune-to-serverless.md) |\n| Interactive dev box (SSH \u002F VS Code) | [06](.\u002Fgolden-paths\u002F06-dev-pod.md) |\n| Move data pod → volume → serverless | [07](.\u002Fgolden-paths\u002F07-network-volume-handoff.md) |\n| **Custom serverless when flash isn't enough** (dual-mode image dev loop) | [09](.\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002FREADME.md) |\n| Build a minimal image for a target (pod vs serverless queue) | [22 (pod)](.\u002Fgolden-paths\u002F22-minimal-pod-image\u002FREADME.md), [23 (queue)](.\u002Fgolden-paths\u002F23-minimal-queue-image\u002FREADME.md); concepts in [building-images](..\u002Frunpod-usage\u002Freference\u002Fbuilding-images.md) |\n| Decide what to bake into the image vs mount on a network volume | [25 — bake vs mount](.\u002Fgolden-paths\u002F25-bake-vs-mount\u002FREADME.md) |\n| Pick a network-volume **storage tier** (standard vs high-performance) | [21 — storage tiers](.\u002Fgolden-paths\u002F21-storage-tiers.md) |\n| **High availability \u002F multi-region** serverless (multi-volume + data sync) | [10](.\u002Fgolden-paths\u002F10-multi-region-ha-serverless.md), [19 (3-region)](.\u002Fgolden-paths\u002F19-three-region-same-file.md) |\n| Stream output incrementally (`\u002Fstream`) | [12](.\u002Fgolden-paths\u002F12-serverless-streaming.md) |\n| Tune autoscaling \u002F raise per-worker throughput | [13 (autoscaling)](.\u002Fgolden-paths\u002F13-autoscaling-tuning.md), [18 (concurrency)](.\u002Fgolden-paths\u002F18-concurrent-handler.md) |\n| Load-balancing \u002F HTTP-server or WebSocket worker | [14 (LB)](.\u002Fgolden-paths\u002F14-load-balancing-endpoint.md), [17 (WebSocket)](.\u002Fgolden-paths\u002F17-serverless-websocket.md) |\n| Get notified on job completion (push, not poll) | [16 — webhooks](.\u002Fgolden-paths\u002F16-serverless-webhooks.md) |\n| Check health \u002F debug a failing endpoint | [15 — monitor & debug](.\u002Fgolden-paths\u002F15-monitor-and-debug.md) |\n\n## Multi-lane tasks\n\nSequence is always **understand → produce artifacts → manage infra → verify**,\nbecause infra can only reference artifacts that already exist. Keep each step in\none lane, and switch lanes at credential boundaries.\n\nExample — \"deploy `openai\u002Fgpt-oss-20b` to a serverless endpoint\":\n1. **runpod-usage** — serverless vs pod, GPU tier for 20B, bake vs mount vs cache.\n2. **companion-clis** — `hf download …`, `docker build --platform=linux\u002Famd64 …`, `docker push`.\n3. **runpod-mcp** or **runpodctl** — create the endpoint referencing the image + GPU pool.\n4. Same infra lane — invoke the endpoint \u002F check status to verify.\n\n## Auth\n\nEverything is one key: **`RUNPOD_API_KEY`** (https:\u002F\u002Fconsole.runpod.io\u002Fuser\u002Fsettings).\nEach lane just makes that key resolvable — `runpodctl doctor`, `flash login`, MCP\nstdio env var, or MCP hosted \"Sign in with Runpod\" (OAuth, no key on disk).\nCompanion CLIs use their **own** credentials (HuggingFace token, GitHub auth,\nDocker Hub PAT, Runpod **S3** keys for `aws`) — do not reuse `RUNPOD_API_KEY` for\nthose.\n",{"name":5,"license":15,"metadata":24,"description":7},{"author":5,"version":25},"1.2.0","main","7540bbccf0c1ded6156a467b02c95b61ad62413f",".agents\u002Fskills\u002Frunpod","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena","Currently exploring the intersection of tech","2026-09-11 03:02:31",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":45},[35,38,41],{"tool":36,"command":37},"npx","npx skills add NeelakshSaxena\u002FVayu --skill runpod",{"tool":39,"command":40},"bunx","bunx skills add NeelakshSaxena\u002FVayu --skill runpod",{"tool":42,"command":43},"pnpm","pnpm dlx skills add NeelakshSaxena\u002FVayu --skill runpod","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002FSKILL.md",true,[47,53,57,61,65,69,73,77,81,85,89,93,97,101,105,109,113,117,121,125,129,133,139,145,149,153,157,161,165,169,173,177,181,185,189,193,197,201,205,209,213,217,221,225,229,233,237,241],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"evals\u002Fconsult-golden-path-first.eval.md","REFERENCE","参考文档",4549,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Fconsult-golden-path-first.eval.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"evals\u002Fmcp-vs-runpodctl-tiebreak.eval.md",1312,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Fmcp-vs-runpodctl-tiebreak.eval.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"evals\u002Fno-unchecked-absence-claims.eval.md",1886,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Fno-unchecked-absence-claims.eval.md",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"evals\u002Fprefer-prebuilt.eval.md",1144,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Fprefer-prebuilt.eval.md",{"path":66,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":67,"sourceHtmlUrl":68},"evals\u002Froute-serverless-observability.eval.md",2856,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Froute-serverless-observability.eval.md",{"path":70,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":71,"sourceHtmlUrl":72},"evals\u002Froute-service-vs-endpoint.eval.md",1154,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Froute-service-vs-endpoint.eval.md",{"path":74,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":75,"sourceHtmlUrl":76},"evals\u002Froute-template-repair.eval.md",1569,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Froute-template-repair.eval.md",{"path":78,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":79,"sourceHtmlUrl":80},"evals\u002Froute-templates.eval.md",1258,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fevals\u002Froute-templates.eval.md",{"path":82,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":83,"sourceHtmlUrl":84},"golden-paths\u002F01-ollama-pod.md",12969,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F01-ollama-pod.md",{"path":86,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":87,"sourceHtmlUrl":88},"golden-paths\u002F02-comfyui-pod\u002FREADME.md",6048,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F02-comfyui-pod\u002FREADME.md",{"path":90,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":91,"sourceHtmlUrl":92},"golden-paths\u002F02-comfyui-pod\u002Fvariant-a-from-scratch.md",5786,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F02-comfyui-pod\u002Fvariant-a-from-scratch.md",{"path":94,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":95,"sourceHtmlUrl":96},"golden-paths\u002F02-comfyui-pod\u002Fvariant-b-prebuilt.md",6479,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F02-comfyui-pod\u002Fvariant-b-prebuilt.md",{"path":98,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":99,"sourceHtmlUrl":100},"golden-paths\u002F03-whisper-endpoint\u002FREADME.md",6362,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F03-whisper-endpoint\u002FREADME.md",{"path":102,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":103,"sourceHtmlUrl":104},"golden-paths\u002F03-whisper-endpoint\u002Fvariant-a-hub.md",6040,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F03-whisper-endpoint\u002Fvariant-a-hub.md",{"path":106,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":107,"sourceHtmlUrl":108},"golden-paths\u002F03-whisper-endpoint\u002Fvariant-b-flash.md",5804,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F03-whisper-endpoint\u002Fvariant-b-flash.md",{"path":110,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":111,"sourceHtmlUrl":112},"golden-paths\u002F04-finetune-pod.md",13985,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F04-finetune-pod.md",{"path":114,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":115,"sourceHtmlUrl":116},"golden-paths\u002F05-model-to-endpoint-pipeline.md",11430,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F05-model-to-endpoint-pipeline.md",{"path":118,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":119,"sourceHtmlUrl":120},"golden-paths\u002F06-dev-pod.md",12249,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F06-dev-pod.md",{"path":122,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":123,"sourceHtmlUrl":124},"golden-paths\u002F07-network-volume-handoff.md",9876,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F07-network-volume-handoff.md",{"path":126,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":127,"sourceHtmlUrl":128},"golden-paths\u002F08-finetune-to-serverless.md",13473,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F08-finetune-to-serverless.md",{"path":130,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":131,"sourceHtmlUrl":132},"golden-paths\u002F09-custom-serverless-dev-loop\u002FREADME.md",18270,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002FREADME.md",{"path":134,"resourceType":135,"resourceTypeDesc":136,"sizeBytes":137,"sourceHtmlUrl":138},"golden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002FDockerfile","OTHER","其他",1711,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002FDockerfile",{"path":140,"resourceType":141,"resourceTypeDesc":142,"sizeBytes":143,"sourceHtmlUrl":144},"golden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002Fhandler.py","SCRIPT","脚本",3701,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002Fhandler.py",{"path":146,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":147,"sourceHtmlUrl":148},"golden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002Frequirements.txt",500,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002Frequirements.txt",{"path":150,"resourceType":141,"resourceTypeDesc":142,"sizeBytes":151,"sourceHtmlUrl":152},"golden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002Fstart.sh",1564,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F09-custom-serverless-dev-loop\u002Ftemplate\u002Fstart.sh",{"path":154,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":155,"sourceHtmlUrl":156},"golden-paths\u002F10-multi-region-ha-serverless.md",22260,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F10-multi-region-ha-serverless.md",{"path":158,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":159,"sourceHtmlUrl":160},"golden-paths\u002F11-public-endpoints.md",11007,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F11-public-endpoints.md",{"path":162,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":163,"sourceHtmlUrl":164},"golden-paths\u002F12-serverless-streaming.md",8848,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F12-serverless-streaming.md",{"path":166,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":167,"sourceHtmlUrl":168},"golden-paths\u002F13-autoscaling-tuning.md",15731,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F13-autoscaling-tuning.md",{"path":170,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":171,"sourceHtmlUrl":172},"golden-paths\u002F14-load-balancing-endpoint.md",15356,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F14-load-balancing-endpoint.md",{"path":174,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":175,"sourceHtmlUrl":176},"golden-paths\u002F15-monitor-and-debug.md",27520,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F15-monitor-and-debug.md",{"path":178,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":179,"sourceHtmlUrl":180},"golden-paths\u002F16-serverless-webhooks.md",10015,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F16-serverless-webhooks.md",{"path":182,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":183,"sourceHtmlUrl":184},"golden-paths\u002F17-serverless-websocket.md",19955,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F17-serverless-websocket.md",{"path":186,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":187,"sourceHtmlUrl":188},"golden-paths\u002F18-concurrent-handler.md",11767,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F18-concurrent-handler.md",{"path":190,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":191,"sourceHtmlUrl":192},"golden-paths\u002F19-three-region-same-file.md",17350,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F19-three-region-same-file.md",{"path":194,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":195,"sourceHtmlUrl":196},"golden-paths\u002F20-model-caching-endpoint.md",9443,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F20-model-caching-endpoint.md",{"path":198,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":199,"sourceHtmlUrl":200},"golden-paths\u002F21-storage-tiers.md",4689,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F21-storage-tiers.md",{"path":202,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":203,"sourceHtmlUrl":204},"golden-paths\u002F22-minimal-pod-image\u002FREADME.md",4979,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F22-minimal-pod-image\u002FREADME.md",{"path":206,"resourceType":135,"resourceTypeDesc":136,"sizeBytes":207,"sourceHtmlUrl":208},"golden-paths\u002F22-minimal-pod-image\u002Ftemplate\u002FDockerfile",803,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F22-minimal-pod-image\u002Ftemplate\u002FDockerfile",{"path":210,"resourceType":141,"resourceTypeDesc":142,"sizeBytes":211,"sourceHtmlUrl":212},"golden-paths\u002F22-minimal-pod-image\u002Ftemplate\u002Fstart.sh",805,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F22-minimal-pod-image\u002Ftemplate\u002Fstart.sh",{"path":214,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":215,"sourceHtmlUrl":216},"golden-paths\u002F23-minimal-queue-image\u002FREADME.md",4952,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F23-minimal-queue-image\u002FREADME.md",{"path":218,"resourceType":135,"resourceTypeDesc":136,"sizeBytes":219,"sourceHtmlUrl":220},"golden-paths\u002F23-minimal-queue-image\u002Ftemplate\u002FDockerfile",436,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F23-minimal-queue-image\u002Ftemplate\u002FDockerfile",{"path":222,"resourceType":141,"resourceTypeDesc":142,"sizeBytes":223,"sourceHtmlUrl":224},"golden-paths\u002F23-minimal-queue-image\u002Ftemplate\u002Fhandler.py",321,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F23-minimal-queue-image\u002Ftemplate\u002Fhandler.py",{"path":226,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":227,"sourceHtmlUrl":228},"golden-paths\u002F23-minimal-queue-image\u002Ftemplate\u002Frequirements.txt",7,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F23-minimal-queue-image\u002Ftemplate\u002Frequirements.txt",{"path":230,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":231,"sourceHtmlUrl":232},"golden-paths\u002F25-bake-vs-mount\u002FREADME.md",6446,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F25-bake-vs-mount\u002FREADME.md",{"path":234,"resourceType":135,"resourceTypeDesc":136,"sizeBytes":235,"sourceHtmlUrl":236},"golden-paths\u002F25-bake-vs-mount\u002Ftemplate\u002FDockerfile",796,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F25-bake-vs-mount\u002Ftemplate\u002FDockerfile",{"path":238,"resourceType":141,"resourceTypeDesc":142,"sizeBytes":239,"sourceHtmlUrl":240},"golden-paths\u002F25-bake-vs-mount\u002Ftemplate\u002Fstart.sh",426,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002F25-bake-vs-mount\u002Ftemplate\u002Fstart.sh",{"path":242,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":243,"sourceHtmlUrl":244},"golden-paths\u002FREADME.md",11094,"https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002Frunpod\u002Fgolden-paths\u002FREADME.md",[246,253,258,265,272,279],{"skillNo":247,"name":248,"nameZh":6,"description":249,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":250,"categories":251,"keywords":252,"publishedAt":19},"01M7ACCCVNCFKMA9X6D15X6G98","runpodctl","Runpod CLI for managing GPU\u002FCPU workloads from the terminal — pods, serverless endpoints, templates, network volumes, Hub deploys, models, SSH, and file transfer (send\u002Freceive). Use for terminal\u002FCI\u002Fscripting, Hub browse\u002Fdeploy, SSH setup, `doctor`, or when the Runpod MCP tools are not connected. For structured tool calls in an MCP-enabled session, prefer runpod-mcp.","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frunpodctl\u002FSKILL.md",[],[],{"skillNo":254,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":255,"categories":256,"keywords":257,"publishedAt":19},"01M7ACC63SRHEA5KMCHSMZK2TC","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frunpod\u002FSKILL.md",[],[],{"skillNo":259,"name":260,"nameZh":6,"description":261,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":262,"categories":263,"keywords":264,"publishedAt":19},"01M7ACC0Q1C89N9BZBV91CTQTF","runpod-usage","How Runpod works and how to work it — pods vs serverless, GPU\u002FVRAM selection, storage, building a container, networking, plus the agentic pod development loop (provision → ssh-exec → set up → poll readiness) and on-pod install hygiene (uv\u002Fapt). Use to answer \"how does X work\", \"which GPU\", \"how do I build a container\", or \"how do I stand up a workload on a pod\". Guidance, not a tool — execute with runpodctl, runpod-mcp, or flash.","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frunpod-usage\u002FSKILL.md",[],[],{"skillNo":266,"name":267,"nameZh":6,"description":268,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":269,"categories":270,"keywords":271,"publishedAt":19},"01M7ACBRPE345GF5GM5B8P8RRC","runpod-templates","Runpod's official prebuilt pod templates (ComfyUI, PyTorch, and others): what each image ships, which ports and paths it uses, how to pick one, how to pin a version, and what is missing on first boot. Use when a task says \"run \u003CX> on a pod\" and an official template already covers it, instead of building an image — and when a user needs help with or wants to fix something about a Runpod template they are already running (won't boot, can't reach the UI, missing models, wrong CUDA line, broken ComfyUI workflow metadata on imported workflows): start here and route onward. Deploy with runpodctl or runpod-mcp; end-to-end walkthroughs live in the runpod router's golden paths.","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frunpod-templates\u002FSKILL.md",[],[],{"skillNo":273,"name":274,"nameZh":6,"description":275,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":276,"categories":277,"keywords":278,"publishedAt":19},"01M7ACBJATHKVX1TWES2F44GY8","runpod-migrate","Migrate a codebase from the Runpod GraphQL API or REST v1 to REST v2 — inventory which parts use which API version, rewrite the call sites, flag breaking changes, and verify. Use when someone asks to move to v2, asks what v2 would change, or asks which Runpod API their code is on. For managing infrastructure rather than migrating code, use runpod-mcp or runpodctl.","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frunpod-migrate\u002FSKILL.md",[],[],{"skillNo":280,"name":281,"nameZh":6,"description":282,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":283,"categories":284,"keywords":285,"publishedAt":19},"01M7ACBC7GHVH6T9YEAPCG2AP8","runpod-mcp","Manage Runpod infrastructure — pods, serverless endpoints, jobs, templates, network volumes, container-registry auth, GPU\u002FCPU catalog, and billing — via the Runpod MCP server's structured tool calls. Use when the Runpod MCP tools (create-pod, list-endpoints, …) are connected in this session, or to connect them (hosted OAuth or local npx). Prefer this over runpodctl for plain infra CRUD when MCP is available; use runpodctl for the terminal, file transfer, or SSH setup.","https:\u002F\u002Fgithub.com\u002FNeelakshSaxena\u002FVayu\u002Fblob\u002Fmain\u002F.claude\u002Fskills\u002Frunpod-mcp\u002FSKILL.md",[],[]]