[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M79111V1PPQNDMSSH6AX65B1":3,"skill-same:01M79111V1PPQNDMSSH6AX65B1":67},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":27,"commitSha":28,"skillDirectory":29,"forksCount":30,"repositoryHtmlUrl":31,"repositoryHomepageUrl":6,"ownerHtmlUrl":32,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":33,"install":34,"resources":46},"01M79111V1PPQNDMSSH6AX65B1","mjx-acceleration",null,"Route optional MyoSuite JAX\u002FMJX acceleration work: probe dependencies and devices, construct MJX environments, use JAX quaternion, fatigue, and reference-motion helpers, and keep CPU substitution and CUDA claims explicit.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: \"mjx-acceleration\"\ndescription: \"Route optional MyoSuite JAX\u002FMJX acceleration work: probe\n  dependencies and devices, construct MJX environments, use JAX quaternion,\n  fatigue, and reference-motion helpers, and keep CPU substitution and CUDA\n  claims explicit.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# MJX acceleration\n\nUse this sub-skill only when the user explicitly asks for the optional JAX\u002FMJX\nor MJWarp route, JAX math parity, JAX fatigue\u002Freference-motion helpers, or\nbackend\u002Fdevice diagnostics. For ordinary MyoSuite environments and MuJoCo\nCPU workflows, route to the environments sub-skill instead. For policy\ntraining recipes, route to training-integration.\n\n## Non-negotiable boundary\n\n- Base MyoSuite is a CPU-capable MuJoCo\u002FGymnasium package.\n- MJX is optional; it is not proof that a CPU-only installation can provide\n  acceleration.\n- CUDA is a separate optional route. Never report a CUDA speedup without a\n  CUDA-backed JAX device probe and a workload measurement.\n- A CPU JAX probe can validate imports, shapes, transforms, and semantics, but\n  it cannot validate CUDA kernels, GPU throughput, or a speedup claim.\n- Do not install packages, initialize assets, download data, or run training\n  from the bundled probe. Installation is an explicit user decision.\n\nRead [compatibility.md](references\u002Fcompatibility.md) before choosing an\ninstallation or device route. Read [api-reference.md](references\u002Fapi-reference.md)\nfor signatures and data contracts. Read [troubleshooting.md](references\u002Ftroubleshooting.md)\nwhen a probe, parity check, or environment construction fails.\n\n## Route selection\n\n1. Run the safe, import-only probe from this directory:\n\n   ```bash\n   python scripts\u002Fmjx_probe.py\n   ```\n\n2. If JAX\u002FMJX is absent, choose one of these honest outcomes:\n   - stay on the base CPU route and state that MJX is optional and unverified;\n   - ask permission to install the `mjx` extra, then rerun the probe;\n   - ask for a CUDA-capable environment and permission to install\n     `mjx-cuda` when GPU acceleration is required.\n\n3. If MJX is present, probe an environment without training:\n\n   ```bash\n   python scripts\u002Fmjx_probe.py --env-name MjxElbowPoseRandom-v0 --steps 1\n   ```\n\n4. For a CUDA claim, require both the optional dependencies and a device whose\n   JAX platform is GPU\u002FCUDA. Use `--require-cuda`; a CPU result is a backend\n   block, not a substitute:\n\n   ```bash\n   python scripts\u002Fmjx_probe.py --require-cuda --env-name MjxElbowPoseRandom-v0\n   ```\n\nThe probe returns JSON only with `--json`, exits nonzero for a requested\nrequired backend that is unavailable, and otherwise leaves optional absence\nas a clearly labelled result.\n\n## Minimal MJX environment contract\n\nThe public constructor is `myosuite.envs.myo.mjx.make(name,\nconfig_overrides=None)`. A normal JAX lifecycle is:\n\n```python\nimport jax\nimport jax.numpy as jnp\nfrom myosuite.envs.myo import mjx\n\nenv = mjx.make(\"MjxElbowPoseRandom-v0\", config_overrides={\"num_envs\": 1})\nstate = env.reset(jax.random.PRNGKey(0))\nnext_state = env.step(state, jnp.zeros((env.action_size,), dtype=jnp.float32))\n```\n\n`reset` consumes a JAX PRNG key and returns a MuJoCo Playground-style `State`.\n`step` returns a new state rather than mutating the old state. The state carries\nMJX data, a dictionary observation, scalar reward\u002Fdone values, metrics, and\ninformation such as the RNG, target, and step count. Confirm action size and\nobservation keys from the concrete environment; do not assume Gymnasium's\nfive-value `step` contract here.\n\nSupported factory names in the inspected implementation are\n`MjxElbowPoseFixed-v0`, `MjxElbowPoseRandom-v0`, `MjxFingerPoseFixed-v0`,\n`MjxFingerPoseRandom-v0`, `MjxHandReachFixed-v0`, and\n`MjxHandReachRandom-v0`. Registering a base environment also exposes fatigue\nvariants with the `MjxFati` prefix. Registration is lazy through `make`; call\n`make` before relying on a name being present in the local registry.\n\n## JAX helper routes\n\n- Quaternion\u002Fmath parity: compare the NumPy helpers in\n  `myosuite.utils.quat_math` with the JAX helpers in\n  `myosuite.utils.quat_math_jax`. Check shape, finite values, rotation-matrix\n  equivalence, and dtype before comparing numerical values.\n- Fatigue: use `CumulativeFatigue.reset(key, ...)` to obtain a state dict with\n  `MA`, `MR`, and `MF`, then pass that state to `compute_act(target_load,\n  state)`. `get_effort` returns the norm of active-minus-target activation.\n- Fatigue wrapper: configure `fatigue_config` with `fatigue_reset_vec`,\n  `fatigue_reset_random`, and allowed observation keys `MA`, `MR`, `MF`.\n  The wrapper stores fatigue state in MJX `userdata` and applies fatigue to\n  muscle activations before stepping.\n- Reference motion: use `ReferenceMotion` with a dict or a supported `.npz`\u002F\n  pickle input. It classifies one-frame, two-bound, and multi-frame data as\n  fixed, random, and track references, respectively; `get_reference(time)`\n  returns a `ReferenceStruct`.\n\n## CPU substitution and evidence\n\nCPU substitution is full for base MuJoCo environments and only partial for\nthis sub-skill. It can validate helper semantics and a CPU JAX execution, but\nnot GPU placement or acceleration. Preserve an explicit `optional-unverified`\nstatus when the extra is not installed. Treat benchmark figures as context\nonly: throughput depends on model contacts, batch size, JAX implementation,\nhardware, and compilation\u002Fwarm-up costs.\n\n## Handoff checklist\n\nBefore claiming success, record:\n\n- package and optional-extra availability;\n- `jax.devices()` and the selected platform;\n- model\u002Fenvironment name, implementation, action shape, and one-step result;\n- parity tolerance, dtype, and quaternion sign handling;\n- whether fatigue\u002Freference-motion helpers were actually exercised;\n- whether the result is CPU semantic validation, CUDA execution, or an\n  unverified optional route.\n\nThe bundled script is deliberately a probe, not a benchmark or installer. Keep\nlong API notes, compatibility rules, and recovery steps in the linked\nreferences.\n","# MJX acceleration\n\nUse this sub-skill only when the user explicitly asks for the optional JAX\u002FMJX\nor MJWarp route, JAX math parity, JAX fatigue\u002Freference-motion helpers, or\nbackend\u002Fdevice diagnostics. For ordinary MyoSuite environments and MuJoCo\nCPU workflows, route to the environments sub-skill instead. For policy\ntraining recipes, route to training-integration.\n\n## Non-negotiable boundary\n\n- Base MyoSuite is a CPU-capable MuJoCo\u002FGymnasium package.\n- MJX is optional; it is not proof that a CPU-only installation can provide\n  acceleration.\n- CUDA is a separate optional route. Never report a CUDA speedup without a\n  CUDA-backed JAX device probe and a workload measurement.\n- A CPU JAX probe can validate imports, shapes, transforms, and semantics, but\n  it cannot validate CUDA kernels, GPU throughput, or a speedup claim.\n- Do not install packages, initialize assets, download data, or run training\n  from the bundled probe. Installation is an explicit user decision.\n\nRead [compatibility.md](references\u002Fcompatibility.md) before choosing an\ninstallation or device route. Read [api-reference.md](references\u002Fapi-reference.md)\nfor signatures and data contracts. Read [troubleshooting.md](references\u002Ftroubleshooting.md)\nwhen a probe, parity check, or environment construction fails.\n\n## Route selection\n\n1. Run the safe, import-only probe from this directory:\n\n   ```bash\n   python scripts\u002Fmjx_probe.py\n   ```\n\n2. If JAX\u002FMJX is absent, choose one of these honest outcomes:\n   - stay on the base CPU route and state that MJX is optional and unverified;\n   - ask permission to install the `mjx` extra, then rerun the probe;\n   - ask for a CUDA-capable environment and permission to install\n     `mjx-cuda` when GPU acceleration is required.\n\n3. If MJX is present, probe an environment without training:\n\n   ```bash\n   python scripts\u002Fmjx_probe.py --env-name MjxElbowPoseRandom-v0 --steps 1\n   ```\n\n4. For a CUDA claim, require both the optional dependencies and a device whose\n   JAX platform is GPU\u002FCUDA. Use `--require-cuda`; a CPU result is a backend\n   block, not a substitute:\n\n   ```bash\n   python scripts\u002Fmjx_probe.py --require-cuda --env-name MjxElbowPoseRandom-v0\n   ```\n\nThe probe returns JSON only with `--json`, exits nonzero for a requested\nrequired backend that is unavailable, and otherwise leaves optional absence\nas a clearly labelled result.\n\n## Minimal MJX environment contract\n\nThe public constructor is `myosuite.envs.myo.mjx.make(name,\nconfig_overrides=None)`. A normal JAX lifecycle is:\n\n```python\nimport jax\nimport jax.numpy as jnp\nfrom myosuite.envs.myo import mjx\n\nenv = mjx.make(\"MjxElbowPoseRandom-v0\", config_overrides={\"num_envs\": 1})\nstate = env.reset(jax.random.PRNGKey(0))\nnext_state = env.step(state, jnp.zeros((env.action_size,), dtype=jnp.float32))\n```\n\n`reset` consumes a JAX PRNG key and returns a MuJoCo Playground-style `State`.\n`step` returns a new state rather than mutating the old state. The state carries\nMJX data, a dictionary observation, scalar reward\u002Fdone values, metrics, and\ninformation such as the RNG, target, and step count. Confirm action size and\nobservation keys from the concrete environment; do not assume Gymnasium's\nfive-value `step` contract here.\n\nSupported factory names in the inspected implementation are\n`MjxElbowPoseFixed-v0`, `MjxElbowPoseRandom-v0`, `MjxFingerPoseFixed-v0`,\n`MjxFingerPoseRandom-v0`, `MjxHandReachFixed-v0`, and\n`MjxHandReachRandom-v0`. Registering a base environment also exposes fatigue\nvariants with the `MjxFati` prefix. Registration is lazy through `make`; call\n`make` before relying on a name being present in the local registry.\n\n## JAX helper routes\n\n- Quaternion\u002Fmath parity: compare the NumPy helpers in\n  `myosuite.utils.quat_math` with the JAX helpers in\n  `myosuite.utils.quat_math_jax`. Check shape, finite values, rotation-matrix\n  equivalence, and dtype before comparing numerical values.\n- Fatigue: use `CumulativeFatigue.reset(key, ...)` to obtain a state dict with\n  `MA`, `MR`, and `MF`, then pass that state to `compute_act(target_load,\n  state)`. `get_effort` returns the norm of active-minus-target activation.\n- Fatigue wrapper: configure `fatigue_config` with `fatigue_reset_vec`,\n  `fatigue_reset_random`, and allowed observation keys `MA`, `MR`, `MF`.\n  The wrapper stores fatigue state in MJX `userdata` and applies fatigue to\n  muscle activations before stepping.\n- Reference motion: use `ReferenceMotion` with a dict or a supported `.npz`\u002F\n  pickle input. It classifies one-frame, two-bound, and multi-frame data as\n  fixed, random, and track references, respectively; `get_reference(time)`\n  returns a `ReferenceStruct`.\n\n## CPU substitution and evidence\n\nCPU substitution is full for base MuJoCo environments and only partial for\nthis sub-skill. It can validate helper semantics and a CPU JAX execution, but\nnot GPU placement or acceleration. Preserve an explicit `optional-unverified`\nstatus when the extra is not installed. Treat benchmark figures as context\nonly: throughput depends on model contacts, batch size, JAX implementation,\nhardware, and compilation\u002Fwarm-up costs.\n\n## Handoff checklist\n\nBefore claiming success, record:\n\n- package and optional-extra availability;\n- `jax.devices()` and the selected platform;\n- model\u002Fenvironment name, implementation, action shape, and one-step result;\n- parity tolerance, dtype, and quaternion sign handling;\n- whether fatigue\u002Freference-motion helpers were actually exercised;\n- whether the result is CPU semantic validation, CUDA execution, or an\n  unverified optional route.\n\nThe bundled script is deliberately a probe, not a benchmark or installer. Keep\nlong API notes, compatibility rules, and recovery steps in the linked\nreferences.\n",{"name":5,"license":23,"metadata":24,"description":7,"disable-model-invocation":26},"Apache 2.0",{"disco-role":25},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":35,"prompt":45,"zipAvailable":26},[36,39,42],{"tool":37,"command":38},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill mjx-acceleration",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill mjx-acceleration",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill mjx-acceleration","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration\u002FSKILL.md",[47,53,57,61],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fapi-reference.md","REFERENCE","参考文档",7583,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration\u002Freferences\u002Fapi-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Fcompatibility.md",5151,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration\u002Freferences\u002Fcompatibility.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Ftroubleshooting.md",9063,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration\u002Freferences\u002Ftroubleshooting.md",{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002Fmjx_probe.py","SCRIPT","脚本",8535,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmyosuite\u002Fsub-skills\u002Fmjx-acceleration\u002Fscripts\u002Fmjx_probe.py",[68,75,82,89,96,103],{"skillNo":69,"name":70,"nameZh":6,"description":71,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":72,"categories":73,"keywords":74,"publishedAt":18},"01M799NYSH1R3RVAHSM2D88Y84","passnet-triton-opt","Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block\u002Fgrid\u002Fwarp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-triton-opt\u002FSKILL.md",[],[],{"skillNo":76,"name":77,"nameZh":6,"description":78,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":18},"01M799NS5APEC0N1FZ7XZVW6BN","passnet-solve","END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-solve\u002FSKILL.md",[],[],{"skillNo":83,"name":84,"nameZh":6,"description":85,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":86,"categories":87,"keywords":88,"publishedAt":18},"01M799NNY9BKSM21V5R7QVTBJX","passnet-skill","PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under .\u002Fpass_dir\u002F, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-skill\u002FSKILL.md",[],[],{"skillNo":90,"name":91,"nameZh":6,"description":92,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":93,"categories":94,"keywords":95,"publishedAt":18},"01M799NHH6NP7Z1RK77KGXAN6S","passnet-pattern-fusion","Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()\u002Freplacement_args()\u002Freplacement_func() files, when a pass \"failed to match\", or when deciding how to fuse multiple kernels\u002Fops.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-pattern-fusion\u002FSKILL.md",[],[],{"skillNo":97,"name":98,"nameZh":6,"description":99,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":100,"categories":101,"keywords":102,"publishedAt":18},"01M799ND2J4WNC2PGAZ0A06NDF","passnet-orchestrate","Round\u002Fbatch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep\u002Frevert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-orchestrate\u002FSKILL.md",[],[],{"skillNo":104,"name":105,"nameZh":6,"description":106,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":107,"categories":108,"keywords":109,"publishedAt":18},"01M799N8SBTMY02ZDG2AZF5ZWE","passnet-feedback","Fast, deterministic feedback for PassNet work: pre-flight pattern\u002Fmatch verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-feedback\u002FSKILL.md",[],[]]