[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7BJP2RGF579RCX1WFD8P31V":3,"skill-same:01M7BJP2RGF579RCX1WFD8P31V":134},{"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":39,"commitSha":69,"skillDirectory":70,"forksCount":14,"repositoryHtmlUrl":71,"repositoryHomepageUrl":72,"ownerHtmlUrl":73,"ownerBlogUrl":6,"ownerDescription":74,"updatedAt":75,"install":76,"resources":89},"01M7BJP2RGF579RCX1WFD8P31V","physical-ai-neural-reconstruction",null,"Router for NVIDIA NuRec\u002FNRE: USDZ rendering, NCore conversion, 3DGS, gRPC sensor sim, PhysicalAI HF datasets. Do NOT use for SimReady or infra setup.","YYC-Cube\u002FYYC3-AI-Agent-Archive","YYC-Cube","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F235006543?v=4","01M79J2PH4SWD2X0825X6PN748","yyc3","01M79K9J2WD44BBY4662XKNZRB",2,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: physical-ai-neural-reconstruction\ndescription: \"Router for NVIDIA NuRec\u002FNRE: USDZ rendering, NCore conversion, 3DGS, gRPC sensor sim, PhysicalAI HF datasets. Do NOT use for SimReady or infra setup.\"\ncategory: development-code\nlicense: Apache-2.0\nversion: \"0.3.0\"\ntools:\n  - Read\n  - Shell\ncompatibility: >-\n  Router skill; downstream sibling skills require Docker, NVIDIA Container\n  Toolkit, GPU, NGC API key, Hugging Face token with PhysicalAI gated\n  licenses accepted, Python 3.10+, and `huggingface_hub`. Optional:\n  CARLA \u002F Isaac Sim 5.1 \u002F AlpaSim for simulator integration over\n  `serve-grpc`.\nmetadata:\n  author: NVIDIA Physical AI\n  tags:\n    - physical-ai\n    - nurec\n    - neural-reconstruction\n    - router\n    - sensor-sim\n  upstream:\n    repo: https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills\n    branch: main\n    skills_dir: .agents\u002Fskills\u002F\n    skills_dir_alias: skills\u002F\n    index_skill: .agents\u002Fskills\u002FSKILL.md\n    index_skill_name: nurec-index\n    sibling_skills:\n      - name: physical-ai-datasets\n        folder: physical-ai-datasets\u002F\n        upstream: https:\u002F\u002Fhuggingface.co\u002Fnvidia\n      - name: ncore\n        folder: ncore\u002F\n        upstream: https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fncore\n      - name: nre\n        folder: nre\u002F\n        upstream: nvcr.io\u002Fnvidia\u002Fnre\u002Fnre\n      - name: asset-harvester\n        folder: asset-harvester\u002F\n        upstream: https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fasset-harvester\n      - name: nurec-fixer\n        folder: nurec-fixer\u002F\n        upstream: https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fharmonizer\n        hf_model: https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FDiffusionHarmonizer\n  upstream_clone_path: \"${PHYSICAL_AI_SKILL_HUB_UPSTREAM_ROOT:-$HOME\u002F.physical-ai-skill-hub\u002Fupstreams}\u002Fnurec-skills\"\n  upstream_override_env: NUREC_SKILLS_UPSTREAM_ROOT\n---\n\n# Physical AI Neural Reconstruction (NuRec) Router\n\n## Purpose\n\nThis is a **thin router** for NVIDIA Neural Reconstruction (NuRec)\nrequests. It points at the upstream `nurec-index` skill at\n`https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills` and its five sibling skills\n(`physical-ai-datasets`, `ncore`, `nre`, `asset-harvester`,\n`nurec-fixer`). Use this skill to:\n\n- Identify which upstream sibling skill answers a NuRec question.\n- Locate, clone, or refresh the canonical `nurec-skills` checkout.\n- Order multi-step NuRec workflows (data → conversion → train →\n  render → cleanup) before opening the upstream recipe.\n\nThe canonical recipes (training, rendering, data conversion, dataset\ndownloads, object harvesting, frame cleanup) live in the upstream\nsibling skills. **Never copy or reconstruct their commands here.**\n\n**Do NOT use this skill for:**\n\n- SimReady packaging of CAD or source meshes → use\n  `omniverse-cad-to-simready`.\n- Generic USD performance tuning unrelated to NuRec → use\n  `omniverse-usd-performance-tuning`.\n- AKS \u002F OSMO \u002F NIM Operator infrastructure setup → use\n  `physical-ai-infrastructure-setup-and-resilient-scaling`.\n\n## When to Use\n\nRead this skill **first** whenever a user mentions any of:\n\n`nurec`, `nurec router`, `nurec index`, `neural reconstruction`,\n`neural reconstruction engine`, `NRE`, `3DGUT`, `3DGRT`, `USDZ`,\n`NCore V4`, `sensor sim`, `novel view synthesis`,\n`PhysicalAI-Autonomous-Vehicles-NuRec`, `PhysicalAI-NuRec-PPISP`,\n`Cosmos-Drive-Dreams`, `asset harvester`, `nurec fixer`,\n`DiffusionHarmonizer`, `harmonizer`, `difix`, `difix3d`, `serve-grpc`,\n`render-grpc`, `warm serve-grpc`, `nre thin client`, `batch_render_rgb`,\n`nurec teardown`, \"where do I start with NuRec\", \"which NuRec skill\nshould I use for X?\".\n\nDecide which upstream sibling skill answers the question, fetch it\n(see [Locate and fetch the upstream skills](#locate-and-fetch-the-upstream-skills)),\nthen follow that skill's body.\n\n## Prerequisites\n\nRouter skill itself has no runtime prerequisites beyond `git` for\nfetching the upstream. Downstream sibling skills require:\n\n- **Docker + NVIDIA Container Toolkit + GPU** — for `nre`, `nre-tools`,\n  and `nurec-fixer` containers\n  (`nvcr.io\u002Fnvidia\u002Fnre\u002Fnre`, `nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools`,\n  `nvcr.io\u002Fnvidia\u002Fcosmos\u002Fcosmos-predict2-container:1.2`).\n- **NGC API key** (`NGC_API_KEY`) — for pulling NGC containers.\n- **Hugging Face token** (`HF_TOKEN`) with the\n  `nvidia\u002FPhysicalAI-*`, `nvidia\u002FDiffusionHarmonizer`, and\n  `nvidia\u002Fasset-harvester` gated licenses **accepted in advance** on\n  Hugging Face.\n- **Python 3.10+** with `huggingface_hub` installed.\n- **(Optional)** CARLA, Isaac Sim 5.1, or AlpaSim for simulator\n  integration over `serve-grpc`.\n\nVerify secrets safely (do not echo values):\n\n```bash\nhf auth whoami\n[ -n \"${HF_TOKEN:-}\" ]      && echo \"HF_TOKEN length=${#HF_TOKEN}\"      || echo \"HF_TOKEN unset\"\n[ -n \"${NGC_API_KEY:-}\" ]   && echo \"NGC_API_KEY length=${#NGC_API_KEY}\" || echo \"NGC_API_KEY unset\"\n```\n\nSee [`references\u002Fsecrets-handling.md`](references\u002Fsecrets-handling.md)\nfor the bash anti-patterns to avoid.\n\n## What is NuRec?\n\n**NuRec** (NVIDIA Omniverse Neural Reconstruction) takes camera, LiDAR,\nradar, or stereo recordings — typically from a self-driving car or a\nrobot — and turns them into a 3D scene you can re-render from any\nviewpoint. Names that come up a lot:\n\n- **NRE** — \"Neural Reconstruction Engine\". NuRec is the product; NRE\n  is the engine that trains and renders. Both route to the upstream\n  `nre` skill.\n- **USDZ** — the file format of a trained scene. A zip archive that\n  Omniverse, Isaac Sim, and CARLA can open.\n- **NCore V4** — the input format NRE consumes. Raw recordings must be\n  converted to NCore V4 before training.\n- **3DGUT \u002F 3DGRT** — the two 3D Gaussian Splatting flavours used\n  internally by NRE. The default Hydra recipe picks one; most users\n  never set it manually.\n\nA typical NuRec project has three stages:\n\n1. **Get the input** — convert your own recording to NCore V4\n   (`ncore`), or download a pre-converted dataset\n   (`physical-ai-datasets`).\n2. **Train the reconstruction** — feed NCore V4 to NRE; out comes a\n   USDZ (`nre`).\n3. **Render new views** — render images, videos, or LiDAR sweeps from\n   the USDZ (`nre`).\n\nProjects that just want to *use* an existing NVIDIA-published scene\nskip step 2.\n\n## Pick a skill\n\nMatch the user's goal in the left column and open the named upstream\nskill on the right. Arrows mean \"do these in order\".\n\n| I want to… | Upstream skill |\n|------------|----------------|\n| Find or download a NuRec dataset NVIDIA has published | `physical-ai-datasets` |\n| Convert my own camera \u002F LiDAR \u002F radar \u002F depth \u002F stereo recording into NCore V4 | `ncore` |\n| Write a new converter for an unsupported sensor setup (drone, RGB-D, ROS 2 bag, COLMAP, ScanNet++) | `ncore` |\n| Train a 3D reconstruction from an NCore clip | `ncore` → `nre` |\n| Generate the extra inputs NRE needs (segmentation masks, depth, ego mask) | `nre` (uses the `nre-tools` container) |\n| Render a USDZ along the original camera positions | `nre` |\n| Render at full resolution \u002F highest quality | `nre` (see \"Quality presets\") |\n| Render along a shifted trajectory (e.g. car moved 3 m left) | `nre` |\n| Render through a server so CARLA \u002F Isaac Sim \u002F AlpaSim \u002F a custom simulator can ask for frames | `nre` (`serve-grpc`) |\n| Render the same USDZ many times back-to-back from Python with minimal per-call latency | `nre` (warm `serve-grpc` + thin Python client \u002F `batch_render_rgb`) |\n| Render LiDAR sweeps (point clouds) from a USDZ | `nre` (`render-grpc --lidar`) |\n| Skip training and just render a NuRec scene NVIDIA already built | `physical-ai-datasets` → `nre` |\n| Extract individual 3D objects (cars, pedestrians) from a driving clip | `asset-harvester` |\n| Add, remove, or replace cars \u002F pedestrians in a NuRec scene | `asset-harvester` → `nre` |\n| Clean up or harmonize rendered frames (ghosting, floaters, flicker, lighting\u002Fshadows) | `nurec-fixer`, **or** `--enable-difix` inside `nre` for inline rendering |\n| Export the scene as a PLY, mesh, depth maps, ego mask, etc. | `nre` |\n| Upgrade an old USDZ so newer NRE versions load it faster | `nre` (`upgrade-artifact`) |\n| Open a USDZ or PLY in a browser viewer | `nre` (`viewer` \u002F `ply_viewer`) |\n| Measure rendering quality (PSNR, SSIM, LPIPS) against ground truth | `nre` (`eval-rendering-metrics`) |\n| Benchmark different reconstruction methods on the same scenes | `physical-ai-datasets` (`PhysicalAI-NuRec-PPISP`) → `nre` |\n| Train on multiple GPUs or on SLURM | `nre` (Workflow D) |\n\n## Common workflows\n\nSix end-to-end workflows are documented in\n[`references\u002Fworkflows.md`](references\u002Fworkflows.md):\n\n- **A.** Make a NuRec scene from your own recording.\n- **B.** Use a NuRec scene NVIDIA has already trained.\n- **C.** Add, remove, or replace 3D objects in a scene.\n- **D.** Clean up rendered frames.\n- **E.** Benchmark reconstruction quality.\n- **F.** Connect NuRec to a simulator.\n\nOpen that file when the user's task spans more than one sibling skill.\n\n## Sibling skills (upstream)\n\n| Name | Upstream folder | What it does |\n|------|-----------------|--------------|\n| `physical-ai-datasets` | `.agents\u002Fskills\u002Fphysical-ai-datasets\u002F` | Catalog and download recipes for every NVIDIA Physical AI dataset on Hugging Face (driving, robotics, manipulation, NuRec scenes, benchmarks). |\n| `ncore` | `.agents\u002Fskills\u002Fncore\u002F` | Converts any sensor recording to NCore V4 (the format NRE needs). Also covers writing a new converter. |\n| `nre` | `.agents\u002Fskills\u002Fnre\u002F` | The Neural Reconstruction Engine itself. Trains, renders (locally, via warm `serve-grpc` + thin Python client \u002F `batch_render_rgb`, or to an external simulator), exports meshes \u002F point clouds \u002F depth, edits actors, evaluates quality. |\n| `asset-harvester` | `.agents\u002Fskills\u002Fasset-harvester\u002F` | Open-source Apache-2.0 pipeline that extracts individual 3D objects from sparse views in a driving clip and saves them as `.ply` Gaussian splats with metadata. |\n| `nurec-fixer` | `.agents\u002Fskills\u002Fnurec-fixer\u002F` | Standalone NVIDIA **DiffusionHarmonizer** workflow — public successor to the older Fixer \u002F Difix3D+ recipes — that cleans rendered frames, harmonizes inserted actors, evaluates PSNR\u002FLPIPS, and optionally fine-tunes the model. |\n\nFor naming overlaps (NRE vs Fixer, ncore vs nre, AV-NuRec vs\nCosmos-Drive-Dreams, NuRec vs SimReady) see\n[`references\u002Fmix-ups.md`](references\u002Fmix-ups.md).\n\n## Locate and fetch the upstream skills\n\nQuick recipe (full version in\n[`references\u002Fupstream-fetch.md`](references\u002Fupstream-fetch.md)):\n\n```bash\nUPSTREAM_ROOT=\"${NUREC_SKILLS_UPSTREAM_ROOT:-${PHYSICAL_AI_SKILL_HUB_UPSTREAM_ROOT:-$HOME\u002F.physical-ai-skill-hub\u002Fupstreams}}\"\nmkdir -p \"$UPSTREAM_ROOT\"\nif [ -d \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.git\" ]; then\n  git -C \"$UPSTREAM_ROOT\u002Fnurec-skills\" fetch --tags\n  git -C \"$UPSTREAM_ROOT\u002Fnurec-skills\" checkout main\n  git -C \"$UPSTREAM_ROOT\u002Fnurec-skills\" pull --ff-only\nelse\n  git clone --depth 1 https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills.git \\\n    \"$UPSTREAM_ROOT\u002Fnurec-skills\"\nfi\ntest -f \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.agents\u002Fskills\u002FSKILL.md\"\n```\n\nThen read the upstream skill before running any mutating command:\n\n```bash\ncat \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.agents\u002Fskills\u002FSKILL.md\"          # router\ncat \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.agents\u002Fskills\u002F\u003Cfolder>\u002FSKILL.md\" # sibling\n```\n\nLocal lookup order (try in order before the upstream clone):\n\n1. `.agents\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (Cursor, Codex, NemoClaw)\n2. `.claude\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (Claude Code)\n3. `.cursor\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (project-scoped)\n4. `~\u002F.cursor\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (personal skills)\n\n## Hard Rules\n\n- Router only — do not duplicate upstream NuRec recipes here. Read\n  the upstream sibling skill body before running any mutating command.\n- Refer to sibling skills by their `name:` (e.g. `nre`), not by repo\n  path. Folder layouts can change; the name is portable.\n- Clone or refresh `https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills` under the\n  shared upstream root\n  (`${NUREC_SKILLS_UPSTREAM_ROOT:-${PHYSICAL_AI_SKILL_HUB_UPSTREAM_ROOT:-$HOME\u002F.physical-ai-skill-hub\u002Fupstreams}}\u002Fnurec-skills`).\n  Do not scan broad developer workspaces such as `~\u002FCodes` or reuse\n  unrelated old clones.\n- `physical-ai-datasets` covers gated Hugging Face datasets. Do not\n  bypass dataset license terms; the user must accept the\n  `PhysicalAI-*` gated licenses on Hugging Face and provide a token\n  before downloading.\n- Asset Harvester runs **before** packaging into a USDZ. Do not call\n  `nre`'s `export-external-assets` on hand-rolled `.ply` files unless\n  the user explicitly asks to skip Asset Harvester.\n- For artifact cleanup, prefer the built-in `--enable-difix` path in\n  `nre`. Route to the standalone `nurec-fixer` only when the user\n  needs the public code\u002Fmodel card, paired evaluation, fine-tuning,\n  or fixes on previously rendered frames.\n- Do not invent NRE \u002F NCore \u002F DiffusionHarmonizer commands from\n  memory. Re-read the upstream sibling skill — versions move fast\n  (NRE `release_26.04` is the current pinned tag).\n- This router does not deploy infrastructure. Route AKS \u002F OSMO \u002F\n  NIM Operator setup to\n  `physical-ai-infrastructure-setup-and-resilient-scaling`.\n\n## Limitations\n\n- **Router only.** This skill never executes mutating NuRec commands.\n  All training, rendering, conversion, and harmonization happens in\n  upstream sibling skills.\n- **Upstream-pinned.** Recipes live in\n  `https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills`, which evolves outside\n  this repo. Stale clones can drift; always `git pull` the upstream\n  before relying on a sibling skill.\n- **Gated content.** `nvidia\u002FPhysicalAI-*`, `nvidia\u002FDiffusionHarmonizer`,\n  and `nvidia\u002Fasset-harvester` require the user to accept license\n  terms on Hugging Face first. The router cannot bypass this.\n- **Heavy footprint.** A complete NuRec workflow can leave 150 GB+\n  on disk. See [`references\u002Fteardown.md`](references\u002Fteardown.md).\n- **NVIDIA-only stack.** Requires an NVIDIA GPU plus the NVIDIA\n  Container Toolkit. AMD \u002F Intel \u002F Apple Silicon are not supported.\n- **Not a SimReady pipeline.** NuRec produces a renderable USDZ from\n  a recording; SimReady packaging of CAD or source meshes is a\n  different pipeline (see `omniverse-cad-to-simready`).\n\n## Troubleshooting\n\n| Error \u002F symptom | Likely cause | Solution |\n|-----------------|--------------|----------|\n| `nurec-skills` clone missing or empty | Upstream not fetched yet | Run the clone block in [Locate and fetch the upstream skills](#locate-and-fetch-the-upstream-skills) |\n| `403`\u002F`401` pulling `nvidia\u002FPhysicalAI-*` from HF | Gated license not accepted, or `HF_TOKEN` unset \u002F wrong scope | Accept the gated license on Hugging Face, then `hf auth login` with a token that has `read` access |\n| `denied: requested access to the resource is denied` from `nvcr.io\u002Fnvidia\u002Fnre\u002F*` | Missing or expired `NGC_API_KEY` | `docker login nvcr.io` with `$oauthtoken` \u002F `NGC_API_KEY`; rotate the key at `org.ngc.nvidia.com\u002Fsetup\u002Fapi-key` if needed |\n| NRE refuses to load a clip (\"not valid NCore V4\") | Recording was not converted | Run the `ncore` skill before invoking `nre` |\n| `serve-grpc` cold-start latency dominates a Python loop | One-shot Docker invocation per render | Use the `nre` warm `serve-grpc` + thin Python client (`batch_render_rgb`) recipe |\n| Output files are owned by `root` after a `docker run` | `-u $(id -u):$(id -g)` was missing | `sudo chown -R \"$(id -u):$(id -g)\" \u003Coutput_dir>`; add the `-u` flag next time |\n| Frames have ghosting \u002F floaters \u002F flicker after rendering | Inline cleanup not enabled | Re-render with `nre --enable-difix`, or post-process with `nurec-fixer` (DiffusionHarmonizer) |\n| Stale skill names (`ncore-data-conversion`, old `nvidia\u002FFixer`) in agent output | Out-of-date cached skill | Update references to `ncore` and `nurec-fixer` (DiffusionHarmonizer); see [`references\u002Fmaintenance.md`](references\u002Fmaintenance.md) |\n| Bash anti-pattern `${HF_TOKEN:+yes}${HF_TOKEN:-no}` echoed token value | Misuse of bash parameter expansion | Rotate the token; use `hf auth whoami` or length-only checks (see [`references\u002Fsecrets-handling.md`](references\u002Fsecrets-handling.md)) |\n\n## Cross-skill teardown\n\nA complete NuRec workflow can leave **150 GB+** on disk between\ncontainer images, model weights, code clones, conda envs, and output\ndirectories. Each sibling skill has its own dedicated `Teardown`\nsection — read them in the order documented in\n[`references\u002Fteardown.md`](references\u002Fteardown.md) when the user no\nlonger needs the workflow.\n\n## Keeping this router up to date\n\nProcedure for adding new sibling skills, renames, or upstream URL\nchanges lives in [`references\u002Fmaintenance.md`](references\u002Fmaintenance.md).\nTreat the upstream `nurec-index` at\n\u003Chttps:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002FSKILL.md>\nas authoritative; this skill mirrors only the picker tables, the\nworkflow ordering, and the upstream fetch recipe.\n","# Physical AI Neural Reconstruction (NuRec) Router\n\n## Purpose\n\nThis is a **thin router** for NVIDIA Neural Reconstruction (NuRec)\nrequests. It points at the upstream `nurec-index` skill at\n`https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills` and its five sibling skills\n(`physical-ai-datasets`, `ncore`, `nre`, `asset-harvester`,\n`nurec-fixer`). Use this skill to:\n\n- Identify which upstream sibling skill answers a NuRec question.\n- Locate, clone, or refresh the canonical `nurec-skills` checkout.\n- Order multi-step NuRec workflows (data → conversion → train →\n  render → cleanup) before opening the upstream recipe.\n\nThe canonical recipes (training, rendering, data conversion, dataset\ndownloads, object harvesting, frame cleanup) live in the upstream\nsibling skills. **Never copy or reconstruct their commands here.**\n\n**Do NOT use this skill for:**\n\n- SimReady packaging of CAD or source meshes → use\n  `omniverse-cad-to-simready`.\n- Generic USD performance tuning unrelated to NuRec → use\n  `omniverse-usd-performance-tuning`.\n- AKS \u002F OSMO \u002F NIM Operator infrastructure setup → use\n  `physical-ai-infrastructure-setup-and-resilient-scaling`.\n\n## When to Use\n\nRead this skill **first** whenever a user mentions any of:\n\n`nurec`, `nurec router`, `nurec index`, `neural reconstruction`,\n`neural reconstruction engine`, `NRE`, `3DGUT`, `3DGRT`, `USDZ`,\n`NCore V4`, `sensor sim`, `novel view synthesis`,\n`PhysicalAI-Autonomous-Vehicles-NuRec`, `PhysicalAI-NuRec-PPISP`,\n`Cosmos-Drive-Dreams`, `asset harvester`, `nurec fixer`,\n`DiffusionHarmonizer`, `harmonizer`, `difix`, `difix3d`, `serve-grpc`,\n`render-grpc`, `warm serve-grpc`, `nre thin client`, `batch_render_rgb`,\n`nurec teardown`, \"where do I start with NuRec\", \"which NuRec skill\nshould I use for X?\".\n\nDecide which upstream sibling skill answers the question, fetch it\n(see [Locate and fetch the upstream skills](#locate-and-fetch-the-upstream-skills)),\nthen follow that skill's body.\n\n## Prerequisites\n\nRouter skill itself has no runtime prerequisites beyond `git` for\nfetching the upstream. Downstream sibling skills require:\n\n- **Docker + NVIDIA Container Toolkit + GPU** — for `nre`, `nre-tools`,\n  and `nurec-fixer` containers\n  (`nvcr.io\u002Fnvidia\u002Fnre\u002Fnre`, `nvcr.io\u002Fnvidia\u002Fnre\u002Fnre-tools`,\n  `nvcr.io\u002Fnvidia\u002Fcosmos\u002Fcosmos-predict2-container:1.2`).\n- **NGC API key** (`NGC_API_KEY`) — for pulling NGC containers.\n- **Hugging Face token** (`HF_TOKEN`) with the\n  `nvidia\u002FPhysicalAI-*`, `nvidia\u002FDiffusionHarmonizer`, and\n  `nvidia\u002Fasset-harvester` gated licenses **accepted in advance** on\n  Hugging Face.\n- **Python 3.10+** with `huggingface_hub` installed.\n- **(Optional)** CARLA, Isaac Sim 5.1, or AlpaSim for simulator\n  integration over `serve-grpc`.\n\nVerify secrets safely (do not echo values):\n\n```bash\nhf auth whoami\n[ -n \"${HF_TOKEN:-}\" ]      && echo \"HF_TOKEN length=${#HF_TOKEN}\"      || echo \"HF_TOKEN unset\"\n[ -n \"${NGC_API_KEY:-}\" ]   && echo \"NGC_API_KEY length=${#NGC_API_KEY}\" || echo \"NGC_API_KEY unset\"\n```\n\nSee [`references\u002Fsecrets-handling.md`](references\u002Fsecrets-handling.md)\nfor the bash anti-patterns to avoid.\n\n## What is NuRec?\n\n**NuRec** (NVIDIA Omniverse Neural Reconstruction) takes camera, LiDAR,\nradar, or stereo recordings — typically from a self-driving car or a\nrobot — and turns them into a 3D scene you can re-render from any\nviewpoint. Names that come up a lot:\n\n- **NRE** — \"Neural Reconstruction Engine\". NuRec is the product; NRE\n  is the engine that trains and renders. Both route to the upstream\n  `nre` skill.\n- **USDZ** — the file format of a trained scene. A zip archive that\n  Omniverse, Isaac Sim, and CARLA can open.\n- **NCore V4** — the input format NRE consumes. Raw recordings must be\n  converted to NCore V4 before training.\n- **3DGUT \u002F 3DGRT** — the two 3D Gaussian Splatting flavours used\n  internally by NRE. The default Hydra recipe picks one; most users\n  never set it manually.\n\nA typical NuRec project has three stages:\n\n1. **Get the input** — convert your own recording to NCore V4\n   (`ncore`), or download a pre-converted dataset\n   (`physical-ai-datasets`).\n2. **Train the reconstruction** — feed NCore V4 to NRE; out comes a\n   USDZ (`nre`).\n3. **Render new views** — render images, videos, or LiDAR sweeps from\n   the USDZ (`nre`).\n\nProjects that just want to *use* an existing NVIDIA-published scene\nskip step 2.\n\n## Pick a skill\n\nMatch the user's goal in the left column and open the named upstream\nskill on the right. Arrows mean \"do these in order\".\n\n| I want to… | Upstream skill |\n|------------|----------------|\n| Find or download a NuRec dataset NVIDIA has published | `physical-ai-datasets` |\n| Convert my own camera \u002F LiDAR \u002F radar \u002F depth \u002F stereo recording into NCore V4 | `ncore` |\n| Write a new converter for an unsupported sensor setup (drone, RGB-D, ROS 2 bag, COLMAP, ScanNet++) | `ncore` |\n| Train a 3D reconstruction from an NCore clip | `ncore` → `nre` |\n| Generate the extra inputs NRE needs (segmentation masks, depth, ego mask) | `nre` (uses the `nre-tools` container) |\n| Render a USDZ along the original camera positions | `nre` |\n| Render at full resolution \u002F highest quality | `nre` (see \"Quality presets\") |\n| Render along a shifted trajectory (e.g. car moved 3 m left) | `nre` |\n| Render through a server so CARLA \u002F Isaac Sim \u002F AlpaSim \u002F a custom simulator can ask for frames | `nre` (`serve-grpc`) |\n| Render the same USDZ many times back-to-back from Python with minimal per-call latency | `nre` (warm `serve-grpc` + thin Python client \u002F `batch_render_rgb`) |\n| Render LiDAR sweeps (point clouds) from a USDZ | `nre` (`render-grpc --lidar`) |\n| Skip training and just render a NuRec scene NVIDIA already built | `physical-ai-datasets` → `nre` |\n| Extract individual 3D objects (cars, pedestrians) from a driving clip | `asset-harvester` |\n| Add, remove, or replace cars \u002F pedestrians in a NuRec scene | `asset-harvester` → `nre` |\n| Clean up or harmonize rendered frames (ghosting, floaters, flicker, lighting\u002Fshadows) | `nurec-fixer`, **or** `--enable-difix` inside `nre` for inline rendering |\n| Export the scene as a PLY, mesh, depth maps, ego mask, etc. | `nre` |\n| Upgrade an old USDZ so newer NRE versions load it faster | `nre` (`upgrade-artifact`) |\n| Open a USDZ or PLY in a browser viewer | `nre` (`viewer` \u002F `ply_viewer`) |\n| Measure rendering quality (PSNR, SSIM, LPIPS) against ground truth | `nre` (`eval-rendering-metrics`) |\n| Benchmark different reconstruction methods on the same scenes | `physical-ai-datasets` (`PhysicalAI-NuRec-PPISP`) → `nre` |\n| Train on multiple GPUs or on SLURM | `nre` (Workflow D) |\n\n## Common workflows\n\nSix end-to-end workflows are documented in\n[`references\u002Fworkflows.md`](references\u002Fworkflows.md):\n\n- **A.** Make a NuRec scene from your own recording.\n- **B.** Use a NuRec scene NVIDIA has already trained.\n- **C.** Add, remove, or replace 3D objects in a scene.\n- **D.** Clean up rendered frames.\n- **E.** Benchmark reconstruction quality.\n- **F.** Connect NuRec to a simulator.\n\nOpen that file when the user's task spans more than one sibling skill.\n\n## Sibling skills (upstream)\n\n| Name | Upstream folder | What it does |\n|------|-----------------|--------------|\n| `physical-ai-datasets` | `.agents\u002Fskills\u002Fphysical-ai-datasets\u002F` | Catalog and download recipes for every NVIDIA Physical AI dataset on Hugging Face (driving, robotics, manipulation, NuRec scenes, benchmarks). |\n| `ncore` | `.agents\u002Fskills\u002Fncore\u002F` | Converts any sensor recording to NCore V4 (the format NRE needs). Also covers writing a new converter. |\n| `nre` | `.agents\u002Fskills\u002Fnre\u002F` | The Neural Reconstruction Engine itself. Trains, renders (locally, via warm `serve-grpc` + thin Python client \u002F `batch_render_rgb`, or to an external simulator), exports meshes \u002F point clouds \u002F depth, edits actors, evaluates quality. |\n| `asset-harvester` | `.agents\u002Fskills\u002Fasset-harvester\u002F` | Open-source Apache-2.0 pipeline that extracts individual 3D objects from sparse views in a driving clip and saves them as `.ply` Gaussian splats with metadata. |\n| `nurec-fixer` | `.agents\u002Fskills\u002Fnurec-fixer\u002F` | Standalone NVIDIA **DiffusionHarmonizer** workflow — public successor to the older Fixer \u002F Difix3D+ recipes — that cleans rendered frames, harmonizes inserted actors, evaluates PSNR\u002FLPIPS, and optionally fine-tunes the model. |\n\nFor naming overlaps (NRE vs Fixer, ncore vs nre, AV-NuRec vs\nCosmos-Drive-Dreams, NuRec vs SimReady) see\n[`references\u002Fmix-ups.md`](references\u002Fmix-ups.md).\n\n## Locate and fetch the upstream skills\n\nQuick recipe (full version in\n[`references\u002Fupstream-fetch.md`](references\u002Fupstream-fetch.md)):\n\n```bash\nUPSTREAM_ROOT=\"${NUREC_SKILLS_UPSTREAM_ROOT:-${PHYSICAL_AI_SKILL_HUB_UPSTREAM_ROOT:-$HOME\u002F.physical-ai-skill-hub\u002Fupstreams}}\"\nmkdir -p \"$UPSTREAM_ROOT\"\nif [ -d \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.git\" ]; then\n  git -C \"$UPSTREAM_ROOT\u002Fnurec-skills\" fetch --tags\n  git -C \"$UPSTREAM_ROOT\u002Fnurec-skills\" checkout main\n  git -C \"$UPSTREAM_ROOT\u002Fnurec-skills\" pull --ff-only\nelse\n  git clone --depth 1 https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills.git \\\n    \"$UPSTREAM_ROOT\u002Fnurec-skills\"\nfi\ntest -f \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.agents\u002Fskills\u002FSKILL.md\"\n```\n\nThen read the upstream skill before running any mutating command:\n\n```bash\ncat \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.agents\u002Fskills\u002FSKILL.md\"          # router\ncat \"$UPSTREAM_ROOT\u002Fnurec-skills\u002F.agents\u002Fskills\u002F\u003Cfolder>\u002FSKILL.md\" # sibling\n```\n\nLocal lookup order (try in order before the upstream clone):\n\n1. `.agents\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (Cursor, Codex, NemoClaw)\n2. `.claude\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (Claude Code)\n3. `.cursor\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (project-scoped)\n4. `~\u002F.cursor\u002Fskills\u002F\u003Cname>\u002FSKILL.md` (personal skills)\n\n## Hard Rules\n\n- Router only — do not duplicate upstream NuRec recipes here. Read\n  the upstream sibling skill body before running any mutating command.\n- Refer to sibling skills by their `name:` (e.g. `nre`), not by repo\n  path. Folder layouts can change; the name is portable.\n- Clone or refresh `https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills` under the\n  shared upstream root\n  (`${NUREC_SKILLS_UPSTREAM_ROOT:-${PHYSICAL_AI_SKILL_HUB_UPSTREAM_ROOT:-$HOME\u002F.physical-ai-skill-hub\u002Fupstreams}}\u002Fnurec-skills`).\n  Do not scan broad developer workspaces such as `~\u002FCodes` or reuse\n  unrelated old clones.\n- `physical-ai-datasets` covers gated Hugging Face datasets. Do not\n  bypass dataset license terms; the user must accept the\n  `PhysicalAI-*` gated licenses on Hugging Face and provide a token\n  before downloading.\n- Asset Harvester runs **before** packaging into a USDZ. Do not call\n  `nre`'s `export-external-assets` on hand-rolled `.ply` files unless\n  the user explicitly asks to skip Asset Harvester.\n- For artifact cleanup, prefer the built-in `--enable-difix` path in\n  `nre`. Route to the standalone `nurec-fixer` only when the user\n  needs the public code\u002Fmodel card, paired evaluation, fine-tuning,\n  or fixes on previously rendered frames.\n- Do not invent NRE \u002F NCore \u002F DiffusionHarmonizer commands from\n  memory. Re-read the upstream sibling skill — versions move fast\n  (NRE `release_26.04` is the current pinned tag).\n- This router does not deploy infrastructure. Route AKS \u002F OSMO \u002F\n  NIM Operator setup to\n  `physical-ai-infrastructure-setup-and-resilient-scaling`.\n\n## Limitations\n\n- **Router only.** This skill never executes mutating NuRec commands.\n  All training, rendering, conversion, and harmonization happens in\n  upstream sibling skills.\n- **Upstream-pinned.** Recipes live in\n  `https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills`, which evolves outside\n  this repo. Stale clones can drift; always `git pull` the upstream\n  before relying on a sibling skill.\n- **Gated content.** `nvidia\u002FPhysicalAI-*`, `nvidia\u002FDiffusionHarmonizer`,\n  and `nvidia\u002Fasset-harvester` require the user to accept license\n  terms on Hugging Face first. The router cannot bypass this.\n- **Heavy footprint.** A complete NuRec workflow can leave 150 GB+\n  on disk. See [`references\u002Fteardown.md`](references\u002Fteardown.md).\n- **NVIDIA-only stack.** Requires an NVIDIA GPU plus the NVIDIA\n  Container Toolkit. AMD \u002F Intel \u002F Apple Silicon are not supported.\n- **Not a SimReady pipeline.** NuRec produces a renderable USDZ from\n  a recording; SimReady packaging of CAD or source meshes is a\n  different pipeline (see `omniverse-cad-to-simready`).\n\n## Troubleshooting\n\n| Error \u002F symptom | Likely cause | Solution |\n|-----------------|--------------|----------|\n| `nurec-skills` clone missing or empty | Upstream not fetched yet | Run the clone block in [Locate and fetch the upstream skills](#locate-and-fetch-the-upstream-skills) |\n| `403`\u002F`401` pulling `nvidia\u002FPhysicalAI-*` from HF | Gated license not accepted, or `HF_TOKEN` unset \u002F wrong scope | Accept the gated license on Hugging Face, then `hf auth login` with a token that has `read` access |\n| `denied: requested access to the resource is denied` from `nvcr.io\u002Fnvidia\u002Fnre\u002F*` | Missing or expired `NGC_API_KEY` | `docker login nvcr.io` with `$oauthtoken` \u002F `NGC_API_KEY`; rotate the key at `org.ngc.nvidia.com\u002Fsetup\u002Fapi-key` if needed |\n| NRE refuses to load a clip (\"not valid NCore V4\") | Recording was not converted | Run the `ncore` skill before invoking `nre` |\n| `serve-grpc` cold-start latency dominates a Python loop | One-shot Docker invocation per render | Use the `nre` warm `serve-grpc` + thin Python client (`batch_render_rgb`) recipe |\n| Output files are owned by `root` after a `docker run` | `-u $(id -u):$(id -g)` was missing | `sudo chown -R \"$(id -u):$(id -g)\" \u003Coutput_dir>`; add the `-u` flag next time |\n| Frames have ghosting \u002F floaters \u002F flicker after rendering | Inline cleanup not enabled | Re-render with `nre --enable-difix`, or post-process with `nurec-fixer` (DiffusionHarmonizer) |\n| Stale skill names (`ncore-data-conversion`, old `nvidia\u002FFixer`) in agent output | Out-of-date cached skill | Update references to `ncore` and `nurec-fixer` (DiffusionHarmonizer); see [`references\u002Fmaintenance.md`](references\u002Fmaintenance.md) |\n| Bash anti-pattern `${HF_TOKEN:+yes}${HF_TOKEN:-no}` echoed token value | Misuse of bash parameter expansion | Rotate the token; use `hf auth whoami` or length-only checks (see [`references\u002Fsecrets-handling.md`](references\u002Fsecrets-handling.md)) |\n\n## Cross-skill teardown\n\nA complete NuRec workflow can leave **150 GB+** on disk between\ncontainer images, model weights, code clones, conda envs, and output\ndirectories. Each sibling skill has its own dedicated `Teardown`\nsection — read them in the order documented in\n[`references\u002Fteardown.md`](references\u002Fteardown.md) when the user no\nlonger needs the workflow.\n\n## Keeping this router up to date\n\nProcedure for adding new sibling skills, renames, or upstream URL\nchanges lives in [`references\u002Fmaintenance.md`](references\u002Fmaintenance.md).\nTreat the upstream `nurec-index` at\n\u003Chttps:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills\u002Fblob\u002Fmain\u002F.agents\u002Fskills\u002FSKILL.md>\nas authoritative; this skill mirrors only the picker tables, the\nworkflow ordering, and the upstream fetch recipe.\n",{"name":5,"tools":24,"license":15,"version":27,"category":28,"metadata":29,"description":7,"compatibility":68},[25,26],"Read","Shell","0.3.0","development-code",{"tags":30,"author":36,"upstream":37,"upstream_clone_path":66,"upstream_override_env":67},[31,32,33,34,35],"physical-ai","nurec","neural-reconstruction","router","sensor-sim","NVIDIA Physical AI",{"repo":38,"branch":39,"skills_dir":40,"index_skill":41,"sibling_skills":42,"index_skill_name":64,"skills_dir_alias":65},"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fnurec-skills","main",".agents\u002Fskills\u002F",".agents\u002Fskills\u002FSKILL.md",[43,47,51,55,59],{"name":44,"folder":45,"upstream":46},"physical-ai-datasets","physical-ai-datasets\u002F","https:\u002F\u002Fhuggingface.co\u002Fnvidia",{"name":48,"folder":49,"upstream":50},"ncore","ncore\u002F","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fncore",{"name":52,"folder":53,"upstream":54},"nre","nre\u002F","nvcr.io\u002Fnvidia\u002Fnre\u002Fnre",{"name":56,"folder":57,"upstream":58},"asset-harvester","asset-harvester\u002F","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fasset-harvester",{"name":60,"folder":61,"hf_model":62,"upstream":63},"nurec-fixer","nurec-fixer\u002F","https:\u002F\u002Fhuggingface.co\u002Fnvidia\u002FDiffusionHarmonizer","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fharmonizer","nurec-index","skills\u002F","${PHYSICAL_AI_SKILL_HUB_UPSTREAM_ROOT:-$HOME\u002F.physical-ai-skill-hub\u002Fupstreams}\u002Fnurec-skills","NUREC_SKILLS_UPSTREAM_ROOT","Router skill; downstream sibling skills require Docker, NVIDIA Container Toolkit, GPU, NGC API key, Hugging Face token with PhysicalAI gated licenses accepted, Python 3.10+, and `huggingface_hub`. Optional: CARLA \u002F Isaac Sim 5.1 \u002F AlpaSim for simulator integration over `serve-grpc`.","d591030ef9fb1bbcebed3454aa5c2dcb8d608f69","skills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive","https:\u002F\u002Fai-agent.yyc3.vip\u002F","https:\u002F\u002Fgithub.com\u002FYYC-Cube","万象归元于云枢 | 深栈智启新纪元","2026-11-13 22:26:14",{"source":8,"skillName":5,"commands":77,"prompt":87,"zipAvailable":88},[78,81,84],{"tool":79,"command":80},"npx","npx skills add YYC-Cube\u002FYYC3-AI-Agent-Archive --skill physical-ai-neural-reconstruction",{"tool":82,"command":83},"bunx","bunx skills add YYC-Cube\u002FYYC3-AI-Agent-Archive --skill physical-ai-neural-reconstruction",{"tool":85,"command":86},"pnpm","pnpm dlx skills add YYC-Cube\u002FYYC3-AI-Agent-Archive --skill physical-ai-neural-reconstruction","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002FSKILL.md",true,[90,96,100,104,108,112,116,120,124,128],{"path":91,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":94,"sourceHtmlUrl":95},"BENCHMARK.md","REFERENCE","参考文档",3090,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002FBENCHMARK.md",{"path":97,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":98,"sourceHtmlUrl":99},"evals\u002Fevals.json",4976,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Fevals\u002Fevals.json",{"path":101,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":102,"sourceHtmlUrl":103},"references\u002Fmaintenance.md",1240,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Freferences\u002Fmaintenance.md",{"path":105,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":106,"sourceHtmlUrl":107},"references\u002Fmix-ups.md",2197,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Freferences\u002Fmix-ups.md",{"path":109,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":110,"sourceHtmlUrl":111},"references\u002Fsecrets-handling.md",1065,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Freferences\u002Fsecrets-handling.md",{"path":113,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":114,"sourceHtmlUrl":115},"references\u002Fteardown.md",1614,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Freferences\u002Fteardown.md",{"path":117,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":118,"sourceHtmlUrl":119},"references\u002Fupstream-fetch.md",2318,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Freferences\u002Fupstream-fetch.md",{"path":121,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":122,"sourceHtmlUrl":123},"references\u002Fworkflows.md",4268,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Freferences\u002Fworkflows.md",{"path":125,"resourceType":92,"resourceTypeDesc":93,"sizeBytes":126,"sourceHtmlUrl":127},"skill-card.md",2998,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Fskill-card.md",{"path":129,"resourceType":130,"resourceTypeDesc":131,"sizeBytes":132,"sourceHtmlUrl":133},"skill.oms.sig","OTHER","其他",5809,"https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Fskills-hub\u002Fai-ml\u002Fnvidia-skills\u002Fplugins\u002Fnvidia-skills\u002Fskills\u002Fphysical-ai-neural-reconstruction\u002Fskill.oms.sig",[135,143,150,157,164,172],{"skillNo":136,"name":137,"nameZh":6,"description":138,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":139,"licenseFileUrl":6,"sourceHtmlUrl":140,"categories":141,"keywords":142,"publishedAt":19},"01M7BTB92RE8MT5YE706QBNDTQ","yingmi-skill","当用户需要查询基金、策略、公告、财经资讯，做资产配置、组合诊断、风险回测、现金流分析，或生成图表、PDF 时，优先使用本 Skill 获取真实数据与可执行能力。","MIT","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Fyingmi-skill\u002FSKILL.md",[],[],{"skillNo":144,"name":145,"nameZh":6,"description":146,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":139,"licenseFileUrl":6,"sourceHtmlUrl":147,"categories":148,"keywords":149,"publishedAt":19},"01M7BT7ZYR091HPJYEFNWVTEJJ","web-access","所有联网操作必须通过此 skill 处理，包括：搜索、网页抓取、登录后操作、网络交互等。","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Fweb-access\u002FSKILL.md",[],[],{"skillNo":151,"name":152,"nameZh":6,"description":153,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":139,"licenseFileUrl":6,"sourceHtmlUrl":154,"categories":155,"keywords":156,"publishedAt":19},"01M7BT4NKS0DNP8A53M1BGGZR2","tencentmap-map-assistant","腾讯位置服务·地图助手 Skill，一句自然语言调用腾讯地图全套能力，无需开发者账号、开箱即用。提供 AI 旅游攻略、地点搜索（含评分\u002F人均\u002F营业时间）、关键词提示、路线规划（驾车\u002F步行\u002F公交\u002F骑行）、地址解析与逆解析、行政区划、IP 定位、距离计算、天气查询，并可将行程或多 POI 渲染成网页地图。涉及找地点、规划路线、旅游行程、查天气、坐标转换等出行场景时使用。","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Ftencentmap-map-assistant\u002FSKILL.md",[],[],{"skillNo":158,"name":159,"nameZh":6,"description":160,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":139,"licenseFileUrl":6,"sourceHtmlUrl":161,"categories":162,"keywords":163,"publishedAt":19},"01M7BSWKV1T5XYYRGV4W4JP88C","skill-scanner","Scan any agent skill for security risks before you install or use it. Powered by Tencent Zhuque Lab A.I.G (AI-Infra-Guard). 100% local static analysis — no file contents or credentials leave your device. Compatible with CodeBuddy, Cursor, Windsurf, Claude Code, OpenClaw and more. Triggers on: 这个 skill 安全吗, skill 安全扫描, 检查 skill 安全, audit skill, scan skill, check skill safety, analyze skill, inspect skill, verify skill, skill security, skill supply chain. Do NOT trigger for general agent usage, full system health checks, project debugging, or normal development.","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Fskill-scanner\u002FSKILL.md",[],[],{"skillNo":165,"name":166,"nameZh":6,"description":167,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":168,"sourceHtmlUrl":169,"categories":170,"keywords":171,"publishedAt":19},"01M7BSWGTCFW7F123S4BM4XY6W","skill-creator","Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends agent capabilities with specialized knowledge, workflows, or tool integrations.","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Fskill-creator\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Fskill-creator\u002FSKILL.md",[],[],{"skillNo":173,"name":174,"nameZh":6,"description":175,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":139,"licenseFileUrl":6,"sourceHtmlUrl":176,"categories":177,"keywords":178,"publishedAt":19},"01M7BSVV4DH8JK2ZMQSCW25VCD","shader-dev","Comprehensive GLSL shader techniques for creating stunning visual effects — ray marching, SDF modeling, fluid simulation, particle systems, procedural generation, lighting, post-processing, and more.","https:\u002F\u002Fgithub.com\u002FYYC-Cube\u002FYYC3-AI-Agent-Archive\u002Fblob\u002Fmain\u002Ftools-hub\u002Fcode-ide\u002Fskills-marketplace\u002Fskills\u002Fshader-dev\u002FSKILL.md",[],[]]