[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7955TM04AMFZ2KH64V46WSH":3,"skill-same:01M7955TM04AMFZ2KH64V46WSH":74},{"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":26,"commitSha":27,"skillDirectory":28,"forksCount":29,"repositoryHtmlUrl":30,"repositoryHomepageUrl":6,"ownerHtmlUrl":31,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":32,"install":33,"resources":45},"01M7955TM04AMFZ2KH64V46WSH","rl-agent-zoo",null,"Implement SMARTS Agent policies, AgentSpec packages, versioned zoo locators, social-agent bubbles, and optional Ray\u002FRLlib integrations without assuming optional training dependencies are installed.","VectorSpaceLab\u002FAREX-Skill","VectorSpaceLab","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F180637102?v=4","01M7538C5WPSJGWF7KHR4EFZ9F","01M7538C6R2GFP53NPXTCZNBRT",351,"MIT","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: rl-agent-zoo\ndescription: \"Implement SMARTS Agent policies, AgentSpec packages, versioned zoo\n  locators, social-agent bubbles, and optional Ray\u002FRLlib integrations without\n  assuming optional training dependencies are installed.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# RL and agent zoo\n\nUse this route when a user needs a SMARTS policy object, an installable agent\npackage, a registered locator, a social agent or bubble, or an RLlib\u002FRay\nadapter. Keep core policy and registry work separate from the simulation\nlifecycle route. Keep benchmark commands and full training runs reference-only.\n\n## Capability and boundary\n\nThis route covers:\n\n- `smarts.core.agent.Agent` policies and the `act(obs, **configs)` contract.\n- `smarts.zoo.agent_spec.AgentSpec`, registry registration, dynamic module\n  loading, and `make`\u002F`make_agent`.\n- Interface\u002Faction\u002Fobservation alignment for an agent specification.\n- `SocialAgentActor`, `BoidAgentActor`, `Bubble`, airlocks, and policy kwargs.\n- The shape of an `RLlibHiWayEnv` configuration and Ray worker distribution.\n- Safe, import-only validation of a locator with\n  [the bundled checker](scripts\u002Fcheck_agent_locator.py).\n- Packaging an inference policy for a zoo or benchmark consumer.\n\nDo not use this route for Gym reset\u002Fstep\u002Fclose basics or general action and\nobservation catalogues; use `simulation-environments`. Use `cli-integrations`\nfor `scl zoo` and `scl benchmark` command details. Do not promise training,\ncheckpoint downloads, benchmark scores, or regression results.\n\n## Fast decision path\n\n1. Decide whether the deliverable is a core Python policy, an installed zoo\n   package, a social\u002Fbubble actor, or an RLlib integration.\n2. Implement and test the policy with a concrete `AgentInterface`. The policy\n   must return an action accepted by that interface's action space.\n3. Wrap the policy and interface in an `AgentSpec`; register a versioned name\n   from an importable module. Run the locator checker before attempting a\n   simulation.\n4. If RLlib is requested, first probe `ray` and `ray.rllib`. They are optional,\n   and are not installed in the prepared core environment. Do not import\n   `RLlibHiWayEnv` as a core smoke test when Ray is absent.\n5. For a benchmark deliverable, package inference code and its model\u002Fdependency\n   contract, then route installation and benchmark invocation to\n   `cli-integrations`.\n\n## Runtime invariants\n\n- A policy subclasses `Agent` and implements `act(self, obs, **configs)`.\n  `Agent.from_function(callable)` is suitable for a tiny stateless policy.\n- `AgentSpec(interface=..., agent_builder=..., agent_params=...)` describes\n  both the simulator-facing interface and the policy constructor. The builder\n  is a class or callable, not an already-built policy instance.\n- `agent_params` are dispatched as no arguments for `None`, positional\n  arguments for a list\u002Ftuple, keyword arguments for a dict, and one argument\n  otherwise. `build_agent()` raises if the builder is absent or not callable.\n- `AgentSpec` is cloudpickle-checked at construction. Keep builders, params,\n  adapters, and model handles serializable when workers will be used.\n- Locators use `module.importable:registered-name-vN` (or `-latest`), where the\n  registered name includes the version suffix. A bare `registered-name-vN`\n  works only after that name is already registered in the current process.\n- A module locator causes SMARTS to import the module and expects that import\n  to execute `register(...)`. Importability is therefore part of the package\n  contract; installing a distribution alone is not proof of registration.\n- `make(locator, **kwargs)` returns an `AgentSpec`; `make_agent(locator,\n  **kwargs)` returns `(Agent, interface)` and applies kwargs to the registered\n  factory. Use `make` when the interface\u002Fspec must be inspected first.\n- Do not use the deprecated observation\u002Faction\u002Freward\u002Finfo adapters for new\n  core code. They remain fields on the current `AgentSpec` for compatibility;\n  RLlib's current adapter path may still exercise them, so keep any legacy use\n  explicit and test it against the selected RLlib release.\n\n## Checks before handing off\n\n- Confirm `AgentInterface` and the policy's raw action agree. Check the\n  environment's per-agent `action_space[agent_id].contains(action)` rather\n  than relying on a model's declared shape.\n- Confirm the policy's observation adapter output agrees with the declared\n  model observation space, including dtype, bounds, keys, and array shape.\n- Import the package module and run\n  `scripts\u002Fcheck_agent_locator.py --locator module:agent-name-v0`.\n  The checker only imports and inspects the registry; it never installs,\n  constructs a model, trains, downloads a checkpoint, or starts a simulator.\n- Build the `AgentSpec` and policy without a scenario first. Use a bounded\n  simulation smoke only when a scenario is supplied and the lifecycle route\n  owns that smoke.\n- For Ray, validate worker configuration and serialization before attempting\n  any long run. The optional framework status is `unverified` until the user\n  supplies a compatible installed stack.\n\n## Progressive references\n\n- [AgentSpec and registry](references\u002Fagent-spec-and-registry.md) — exact\n  interfaces, locator grammar, registration and construction patterns.\n- [RL frameworks](references\u002Frl-frameworks.md) — RLlibHiWayEnv, Ray workers,\n  spaces, optional Torch\u002FTensorFlow, and resume caveats.\n- [Social agents and bubbles](references\u002Fsocial-agents-and-bubbles.md) — actor,\n  boid, bubble, airlock, travel, and policy-kwargs rules.\n- [Benchmark packaging](references\u002Fbenchmark-packaging.md) — inference package\n  layout, dependency\u002Fmodel contracts, and benchmark compatibility gates.\n- [Troubleshooting](references\u002Ftroubleshooting.md) — installation\u002Fimport,\n  optional dependencies, config\u002Fdata, API\u002FCLI mistakes, and workflow failures.\n","# RL and agent zoo\n\nUse this route when a user needs a SMARTS policy object, an installable agent\npackage, a registered locator, a social agent or bubble, or an RLlib\u002FRay\nadapter. Keep core policy and registry work separate from the simulation\nlifecycle route. Keep benchmark commands and full training runs reference-only.\n\n## Capability and boundary\n\nThis route covers:\n\n- `smarts.core.agent.Agent` policies and the `act(obs, **configs)` contract.\n- `smarts.zoo.agent_spec.AgentSpec`, registry registration, dynamic module\n  loading, and `make`\u002F`make_agent`.\n- Interface\u002Faction\u002Fobservation alignment for an agent specification.\n- `SocialAgentActor`, `BoidAgentActor`, `Bubble`, airlocks, and policy kwargs.\n- The shape of an `RLlibHiWayEnv` configuration and Ray worker distribution.\n- Safe, import-only validation of a locator with\n  [the bundled checker](scripts\u002Fcheck_agent_locator.py).\n- Packaging an inference policy for a zoo or benchmark consumer.\n\nDo not use this route for Gym reset\u002Fstep\u002Fclose basics or general action and\nobservation catalogues; use `simulation-environments`. Use `cli-integrations`\nfor `scl zoo` and `scl benchmark` command details. Do not promise training,\ncheckpoint downloads, benchmark scores, or regression results.\n\n## Fast decision path\n\n1. Decide whether the deliverable is a core Python policy, an installed zoo\n   package, a social\u002Fbubble actor, or an RLlib integration.\n2. Implement and test the policy with a concrete `AgentInterface`. The policy\n   must return an action accepted by that interface's action space.\n3. Wrap the policy and interface in an `AgentSpec`; register a versioned name\n   from an importable module. Run the locator checker before attempting a\n   simulation.\n4. If RLlib is requested, first probe `ray` and `ray.rllib`. They are optional,\n   and are not installed in the prepared core environment. Do not import\n   `RLlibHiWayEnv` as a core smoke test when Ray is absent.\n5. For a benchmark deliverable, package inference code and its model\u002Fdependency\n   contract, then route installation and benchmark invocation to\n   `cli-integrations`.\n\n## Runtime invariants\n\n- A policy subclasses `Agent` and implements `act(self, obs, **configs)`.\n  `Agent.from_function(callable)` is suitable for a tiny stateless policy.\n- `AgentSpec(interface=..., agent_builder=..., agent_params=...)` describes\n  both the simulator-facing interface and the policy constructor. The builder\n  is a class or callable, not an already-built policy instance.\n- `agent_params` are dispatched as no arguments for `None`, positional\n  arguments for a list\u002Ftuple, keyword arguments for a dict, and one argument\n  otherwise. `build_agent()` raises if the builder is absent or not callable.\n- `AgentSpec` is cloudpickle-checked at construction. Keep builders, params,\n  adapters, and model handles serializable when workers will be used.\n- Locators use `module.importable:registered-name-vN` (or `-latest`), where the\n  registered name includes the version suffix. A bare `registered-name-vN`\n  works only after that name is already registered in the current process.\n- A module locator causes SMARTS to import the module and expects that import\n  to execute `register(...)`. Importability is therefore part of the package\n  contract; installing a distribution alone is not proof of registration.\n- `make(locator, **kwargs)` returns an `AgentSpec`; `make_agent(locator,\n  **kwargs)` returns `(Agent, interface)` and applies kwargs to the registered\n  factory. Use `make` when the interface\u002Fspec must be inspected first.\n- Do not use the deprecated observation\u002Faction\u002Freward\u002Finfo adapters for new\n  core code. They remain fields on the current `AgentSpec` for compatibility;\n  RLlib's current adapter path may still exercise them, so keep any legacy use\n  explicit and test it against the selected RLlib release.\n\n## Checks before handing off\n\n- Confirm `AgentInterface` and the policy's raw action agree. Check the\n  environment's per-agent `action_space[agent_id].contains(action)` rather\n  than relying on a model's declared shape.\n- Confirm the policy's observation adapter output agrees with the declared\n  model observation space, including dtype, bounds, keys, and array shape.\n- Import the package module and run\n  `scripts\u002Fcheck_agent_locator.py --locator module:agent-name-v0`.\n  The checker only imports and inspects the registry; it never installs,\n  constructs a model, trains, downloads a checkpoint, or starts a simulator.\n- Build the `AgentSpec` and policy without a scenario first. Use a bounded\n  simulation smoke only when a scenario is supplied and the lifecycle route\n  owns that smoke.\n- For Ray, validate worker configuration and serialization before attempting\n  any long run. The optional framework status is `unverified` until the user\n  supplies a compatible installed stack.\n\n## Progressive references\n\n- [AgentSpec and registry](references\u002Fagent-spec-and-registry.md) — exact\n  interfaces, locator grammar, registration and construction patterns.\n- [RL frameworks](references\u002Frl-frameworks.md) — RLlibHiWayEnv, Ray workers,\n  spaces, optional Torch\u002FTensorFlow, and resume caveats.\n- [Social agents and bubbles](references\u002Fsocial-agents-and-bubbles.md) — actor,\n  boid, bubble, airlock, travel, and policy-kwargs rules.\n- [Benchmark packaging](references\u002Fbenchmark-packaging.md) — inference package\n  layout, dependency\u002Fmodel contracts, and benchmark compatibility gates.\n- [Troubleshooting](references\u002Ftroubleshooting.md) — installation\u002Fimport,\n  optional dependencies, config\u002Fdata, API\u002FCLI mistakes, and workflow failures.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo",27,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab","2026-09-03 10:25:21",{"source":8,"skillName":5,"commands":34,"prompt":44,"zipAvailable":25},[35,38,41],{"tool":36,"command":37},"npx","npx skills add VectorSpaceLab\u002FAREX-Skill --skill rl-agent-zoo",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill rl-agent-zoo",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill rl-agent-zoo","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002FSKILL.md",[46,52,56,60,64,68],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fagent-spec-and-registry.md","REFERENCE","参考文档",5586,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002Freferences\u002Fagent-spec-and-registry.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fbenchmark-packaging.md",3966,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002Freferences\u002Fbenchmark-packaging.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Frl-frameworks.md",6455,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002Freferences\u002Frl-frameworks.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Fsocial-agents-and-bubbles.md",4436,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002Freferences\u002Fsocial-agents-and-bubbles.md",{"path":65,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":66,"sourceHtmlUrl":67},"references\u002Ftroubleshooting.md",5866,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002Freferences\u002Ftroubleshooting.md",{"path":69,"resourceType":70,"resourceTypeDesc":71,"sizeBytes":72,"sourceHtmlUrl":73},"scripts\u002Fcheck_agent_locator.py","SCRIPT","脚本",3659,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fsmarts\u002Fsub-skills\u002Frl-agent-zoo\u002Fscripts\u002Fcheck_agent_locator.py",[75,83,90,97,104,111],{"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":79,"licenseFileUrl":6,"sourceHtmlUrl":80,"categories":81,"keywords":82,"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.","Apache-2.0","https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Ftask-oriented\u002FPassNet\u002Fskills\u002Fpassnet-triton-opt\u002FSKILL.md",[],[],{"skillNo":84,"name":85,"nameZh":6,"description":86,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":79,"licenseFileUrl":6,"sourceHtmlUrl":87,"categories":88,"keywords":89,"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":91,"name":92,"nameZh":6,"description":93,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":79,"licenseFileUrl":6,"sourceHtmlUrl":94,"categories":95,"keywords":96,"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":98,"name":99,"nameZh":6,"description":100,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":79,"licenseFileUrl":6,"sourceHtmlUrl":101,"categories":102,"keywords":103,"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":105,"name":106,"nameZh":6,"description":107,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":79,"licenseFileUrl":6,"sourceHtmlUrl":108,"categories":109,"keywords":110,"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":112,"name":113,"nameZh":6,"description":114,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":79,"licenseFileUrl":6,"sourceHtmlUrl":115,"categories":116,"keywords":117,"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",[],[]]