[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78ZZCGE6NTYEXSW10X9A9D9":3,"skill-same:01M78ZZCGE6NTYEXSW10X9A9D9":79},{"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},"01M78ZZCGE6NTYEXSW10X9A9D9","mctx",null,"Use Mctx for JAX-native Monte Carlo tree search, MuZero\u002FGumbel MuZero policies, stochastic MuZero search, and search-tree inspection.","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\u002Fmctx\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: mctx\ndescription: \"Use Mctx for JAX-native Monte Carlo tree search, MuZero\u002FGumbel\n  MuZero policies, stochastic MuZero search, and search-tree inspection.\"\nmetadata:\n  disco-role: operating\ndisable-model-invocation: true\nlicense: Apache 2.0\n---\n\n# Mctx\n\nMctx is a compact JAX library for batched Monte Carlo tree search. Use this\nskill when the task asks you to plan with a learned model, improve a policy\nwith search, or inspect the resulting search tree.\n\n## Start here\n\n- Install the library with `pip install mctx`.\n- If you want a quick health check, run `python scripts\u002Fcheck_install.py`.\n- If you want a safe runnable example, run\n  `python scripts\u002Fpolicy_improvement_demo.py --num_runs=1`.\n- Read `references\u002Fapi-reference.md` when you need exact function signatures or\n  tensor-shape contracts.\n- Read `references\u002Fworkflows.md` when you need an end-to-end recipe.\n- Read `references\u002Ftroubleshooting.md` when search output, shapes, or optional\n  dependencies are failing.\n\n## What this skill covers\n\n### 1. Batched search on a learned model\nUse `search(...)` when you already have:\n\n- a batched root state,\n- a recurrent model for expansion,\n- a root action-selection function,\n- an interior action-selection function,\n- and a simulation budget.\n\nThis is the lowest-level entry point and is best when you need to customize the\nsearch loop.\n\n### 2. Classic MuZero-style planning\nUse `muzero_policy(...)` when you want the standard MuZero policy improvement\nloop with Dirichlet root noise, PUCT-style action selection, and visit-count\nbased action weights.\n\n### 3. Gumbel MuZero planning\nUse `gumbel_muzero_policy(...)` when you want the Gumbel MuZero variant and\nits default Q-value completion strategy. This is the best default when the task\nmentions policy improvement by planning with Gumbel.\n\n### 4. Stochastic MuZero planning\nUse `stochastic_muzero_policy(...)` when the environment alternates between\ndecision nodes and chance nodes. Provide separate decision and chance recurrent\nfunctions; the wrapper handles the internal state structure.\n\n## How to choose the right helper\n\n- Use `qtransform_by_parent_and_siblings` for the default MuZero-style value\n  normalization.\n- Use `qtransform_completed_by_mix_value` for the default Gumbel MuZero path.\n- Use `qtransform_by_min_max` when you already know a bounded Q-value range.\n- Prefer the policy helper that matches the search style instead of reaching\n  into `_src` internals.\n\n## Inspecting results\n\nAfter search, prefer the public summary APIs over raw tree fields unless you are\ntroubleshooting.\n\n- `PolicyOutput.action` is the chosen action.\n- `PolicyOutput.action_weights` are the training targets for the policy head.\n- `PolicyOutput.search_tree` holds the final `Tree`.\n- `Tree.summary()` exposes root visit counts, visit probabilities, value, and\n  Q-values.\n- `Tree.qvalues(node_index)` is the best way to inspect a node.\n\n## Common input contract\n\n- Treat action masks as `1 = invalid`, `0 = valid`.\n- Keep root and recurrent outputs batched.\n- Make `prior_logits` and `invalid_actions` have the same trailing action\n  dimension.\n- Keep `recurrent_fn` outputs aligned with the batch size of the root.\n- Use `max_depth` only when you intentionally want to cap traversal depth.\n\n## Bundled scripts\n\n- `scripts\u002Fcheck_install.py` verifies the installed package, imports the public\n  API, and reports the active JAX backend.\n- `scripts\u002Fpolicy_improvement_demo.py` is a tiny bandit-style smoke demo that\n  shows how planning can improve a policy.\n\n## Optional dependency note\n\nThis skill does not bundle tree visualization. If you need Graphviz-style tree\nrendering later, treat it as a separate dependency problem instead of assuming\nit is available in a minimal install.\n\n## When to read the deeper references\n\n- Read `references\u002Fapi-reference.md` before wiring a custom model or action\n  mask.\n- Read `references\u002Fworkflows.md` before writing a new MuZero or Gumbel MuZero\n  driver.\n- Read `references\u002Ftroubleshooting.md` when output is flat, invalid-action\n  handling looks wrong, or JAX is on the wrong backend.\n- Read `references\u002Frepo-provenance.md` when you need to check whether this\n  skill is stale against the source checkout that produced it.\n","# Mctx\n\nMctx is a compact JAX library for batched Monte Carlo tree search. Use this\nskill when the task asks you to plan with a learned model, improve a policy\nwith search, or inspect the resulting search tree.\n\n## Start here\n\n- Install the library with `pip install mctx`.\n- If you want a quick health check, run `python scripts\u002Fcheck_install.py`.\n- If you want a safe runnable example, run\n  `python scripts\u002Fpolicy_improvement_demo.py --num_runs=1`.\n- Read `references\u002Fapi-reference.md` when you need exact function signatures or\n  tensor-shape contracts.\n- Read `references\u002Fworkflows.md` when you need an end-to-end recipe.\n- Read `references\u002Ftroubleshooting.md` when search output, shapes, or optional\n  dependencies are failing.\n\n## What this skill covers\n\n### 1. Batched search on a learned model\nUse `search(...)` when you already have:\n\n- a batched root state,\n- a recurrent model for expansion,\n- a root action-selection function,\n- an interior action-selection function,\n- and a simulation budget.\n\nThis is the lowest-level entry point and is best when you need to customize the\nsearch loop.\n\n### 2. Classic MuZero-style planning\nUse `muzero_policy(...)` when you want the standard MuZero policy improvement\nloop with Dirichlet root noise, PUCT-style action selection, and visit-count\nbased action weights.\n\n### 3. Gumbel MuZero planning\nUse `gumbel_muzero_policy(...)` when you want the Gumbel MuZero variant and\nits default Q-value completion strategy. This is the best default when the task\nmentions policy improvement by planning with Gumbel.\n\n### 4. Stochastic MuZero planning\nUse `stochastic_muzero_policy(...)` when the environment alternates between\ndecision nodes and chance nodes. Provide separate decision and chance recurrent\nfunctions; the wrapper handles the internal state structure.\n\n## How to choose the right helper\n\n- Use `qtransform_by_parent_and_siblings` for the default MuZero-style value\n  normalization.\n- Use `qtransform_completed_by_mix_value` for the default Gumbel MuZero path.\n- Use `qtransform_by_min_max` when you already know a bounded Q-value range.\n- Prefer the policy helper that matches the search style instead of reaching\n  into `_src` internals.\n\n## Inspecting results\n\nAfter search, prefer the public summary APIs over raw tree fields unless you are\ntroubleshooting.\n\n- `PolicyOutput.action` is the chosen action.\n- `PolicyOutput.action_weights` are the training targets for the policy head.\n- `PolicyOutput.search_tree` holds the final `Tree`.\n- `Tree.summary()` exposes root visit counts, visit probabilities, value, and\n  Q-values.\n- `Tree.qvalues(node_index)` is the best way to inspect a node.\n\n## Common input contract\n\n- Treat action masks as `1 = invalid`, `0 = valid`.\n- Keep root and recurrent outputs batched.\n- Make `prior_logits` and `invalid_actions` have the same trailing action\n  dimension.\n- Keep `recurrent_fn` outputs aligned with the batch size of the root.\n- Use `max_depth` only when you intentionally want to cap traversal depth.\n\n## Bundled scripts\n\n- `scripts\u002Fcheck_install.py` verifies the installed package, imports the public\n  API, and reports the active JAX backend.\n- `scripts\u002Fpolicy_improvement_demo.py` is a tiny bandit-style smoke demo that\n  shows how planning can improve a policy.\n\n## Optional dependency note\n\nThis skill does not bundle tree visualization. If you need Graphviz-style tree\nrendering later, treat it as a separate dependency problem instead of assuming\nit is available in a minimal install.\n\n## When to read the deeper references\n\n- Read `references\u002Fapi-reference.md` before wiring a custom model or action\n  mask.\n- Read `references\u002Fworkflows.md` before writing a new MuZero or Gumbel MuZero\n  driver.\n- Read `references\u002Ftroubleshooting.md` when output is flat, invalid-action\n  handling looks wrong, or JAX is on the wrong backend.\n- Read `references\u002Frepo-provenance.md` when you need to check whether this\n  skill is stale against the source checkout that produced it.\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\u002Fmctx",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 mctx",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill mctx",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill mctx","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002FSKILL.md",[47,53,57,61,65,69,75],{"path":48,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":51,"sourceHtmlUrl":52},"references\u002Fapi-reference.md","REFERENCE","参考文档",3879,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Freferences\u002Fapi-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Frepo-provenance.md",1340,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Freferences\u002Frepo-provenance.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Frepo-routing-metadata.json",434,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Freferences\u002Frepo-routing-metadata.json",{"path":62,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":63,"sourceHtmlUrl":64},"references\u002Ftroubleshooting.md",2960,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Freferences\u002Ftroubleshooting.md",{"path":66,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":67,"sourceHtmlUrl":68},"references\u002Fworkflows.md",3026,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Freferences\u002Fworkflows.md",{"path":70,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":73,"sourceHtmlUrl":74},"scripts\u002Fcheck_install.py","SCRIPT","脚本",822,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Fscripts\u002Fcheck_install.py",{"path":76,"resourceType":71,"resourceTypeDesc":72,"sizeBytes":77,"sourceHtmlUrl":78},"scripts\u002Fpolicy_improvement_demo.py",4476,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmctx\u002Fscripts\u002Fpolicy_improvement_demo.py",[80,87,94,101,108,115],{"skillNo":81,"name":82,"nameZh":6,"description":83,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":84,"categories":85,"keywords":86,"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":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":93,"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":95,"name":96,"nameZh":6,"description":97,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":98,"categories":99,"keywords":100,"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":102,"name":103,"nameZh":6,"description":104,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":105,"categories":106,"keywords":107,"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":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"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":116,"name":117,"nameZh":6,"description":118,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":119,"categories":120,"keywords":121,"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",[],[]]