[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78ZXD1CR2FAS34PKA4FN22R":3,"skill-same:01M78ZXD1CR2FAS34PKA4FN22R":70},{"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},"01M78ZXD1CR2FAS34PKA4FN22R","evaluation",null,"Evaluate MatterGen crystal structures with MatterSim or precomputed energies, matching ordered or disordered structures against compatible reference datasets and exporting aggregate, detailed, and relaxed-structure results.","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\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: evaluation\ndescription: \"Evaluate MatterGen crystal structures with MatterSim or\n  precomputed energies, matching ordered or disordered structures against\n  compatible reference datasets and exporting aggregate, detailed, and\n  relaxed-structure results.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: MIT\n---\n\n# MatterGen evaluation\n\nUse this sub-skill when the task is to score generated crystal structures for\nvalidity, uniqueness, novelty, stability, energy above hull, relaxation RMSD,\nor related metrics. It covers MatterGen 1.0.3's evaluation API and CLI. It\ndoes not generate structures, run DFT, or acquire large model\u002Freference assets.\n\n## Start with the mode decision\n\nChoose exactly one mode before loading data:\n\n- **Relaxed evaluation** (`relax=True`): pass structures to MatterSim's\n  `BatchRelaxer`; its final structures and total energies are evaluated. Do not\n  pass `energies`\u002F`--energies_path` in this mode.\n- **Precomputed-energy evaluation** (`relax=False`): pass already relaxed\n  structures and one externally computed total energy per structure. The array\n  must have the same count and order as the loader result.\n\nThe implementation raises if both `relax=True` and energies are supplied, even\nthough an older docstring says energies would be ignored. Use the safe wrapper\nat [`scripts\u002Fevaluate_structures.py`](scripts\u002Fevaluate_structures.py) for\ninput\u002Fmode validation and clear errors; it only invokes package APIs when the\nscript is explicitly run.\n\nRead the applicable reference before running:\n\n- [`references\u002Fworkflows.md`](references\u002Fworkflows.md) for recipes, output\n  precedence, and expensive\u002Fnetwork boundaries.\n- [`references\u002Fapi-reference.md`](references\u002Fapi-reference.md) for exact\n  Python and CLI parameters, defaults, and metrics.\n- [`references\u002Fdata-formats.md`](references\u002Fdata-formats.md) for structure,\n  energy-array, reference, and output formats.\n- [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md) for\n  install\u002Fbackend, data, matcher, correction, and workflow failures.\n\n## Minimal preflight\n\n1. Confirm the input path exists and is one of `.xyz`, `.extxyz`, `.zip`, or a\n   directory containing lowercase `.cif`, `.xyz`, or `.extxyz` files.\n2. Record the loader order. A direct XYZ\u002FEXTXYZ file is read frame-by-frame;\n   a ZIP is extracted temporarily; a directory is traversed in the exact\n   `os.listdir` order (not sorted). Keep any energy array in that same order.\n3. For `relax=False`, load a one-dimensional, numeric, finite `energies.npy`\n   and verify `len(energies) == len(structures)` before calling `evaluate`.\n4. Select a matcher: `disordered` is the robust default for generated alloys,\n   partial occupancies, and mixed ordered\u002Fdisordered sets; use `ordered` only\n   when the data and reference are genuinely ordered and strict reduced-formula\n   grouping is intended.\n5. Select a correction\u002Freference pair. MP2020 is the historical benchmark\n   pairing. TRI2024 (`TRI110Compatibility2024`) plus the TRI2024 reference is\n   recommended for work outside historical benchmarking. Never mix a\n   correction scheme with a reference corrected under another scheme.\n6. Check device, MatterSim availability\u002Fcache, disk space for outputs, and\n   whether a large relaxation is affordable. The wrapper does not download\n   datasets or potential files.\n\n## Canonical API\n\nThe package function is:\n\n```python\nevaluate(\n    structures, relax=True, energies=None, reference=None,\n    structure_matcher=DefaultDisorderedStructureMatcher(),\n    save_as=None, save_detailed_as=None, potential_load_path=None,\n    device=\"cuda\", structures_output_path=None,\n    energy_correction_scheme=MaterialsProject2020Compatibility(),\n)\n```\n\nIt returns a dictionary of aggregate metric values. With `reference=None`, the\nmetric evaluator selects the MP2020 preset, or the TRI2024 preset when the\nTRI2024 correction object is supplied. A supplied reference is deserialized by\nthe CLI from the gzip-compressed LMDB format. Missing terminal systems can make\nenergy metrics unavailable while structure metrics remain computable; treat\nwarnings and absent metrics as results to investigate, not as zeros.\n\n## CLI routing\n\nThe safe script uses these exact underscore flags:\n\n```text\n--structures_path PATH                 required\n--relax {True,False}                   default True\n--energies_path PATH                   required in no-relax mode\n--structure_matcher {ordered,disordered}\n--save_as PATH                         aggregate JSON\n--save_detailed_as PATH                per-structure serialized table\n--potential_load_path PATH             local MatterSim checkpoint\n--reference_dataset_path PATH          .gz LMDB reference\n--device DEVICE                        auto, cpu, cuda[:N], or mps\n--structures_output_path PATH          relaxed EXTXYZ, relax mode only\n--energy_correction_scheme {MP2020,TRI2024}\n```\n\nExample relaxed evaluation:\n\n```bash\npython \u003Cmattergen-skill-root>\u002Fsub-skills\u002Fevaluation\u002Fscripts\u002Fevaluate_structures.py \\\n  --structures_path results \\\n  --relax True --structure_matcher disordered \\\n  --reference_dataset_path data\u002Freference_TRI2024correction.gz \\\n  --energy_correction_scheme TRI2024 \\\n  --save_as results\u002Fmetrics.json \\\n  --save_detailed_as results\u002Fdetailed_metrics.json \\\n  --structures_output_path results\u002Frelaxed.extxyz\n```\n\nExample with external energies:\n\n```bash\npython \u003Cmattergen-skill-root>\u002Fsub-skills\u002Fevaluation\u002Fscripts\u002Fevaluate_structures.py \\\n  --structures_path results.extxyz --relax False \\\n  --energies_path energies.npy --structure_matcher disordered \\\n  --reference_dataset_path data\u002Freference_TRI2024correction.gz \\\n  --energy_correction_scheme TRI2024 --save_as metrics.json\n```\n\n## Interpret the result\n\nThe evaluator can report structure metrics (`precision`, `recall`,\n`frac_unique_structures`, `frac_novel_structures`, `frac_novel_unique_structures`,\nchemical-system novelty, validity, and composition validity), energy metrics\n(`avg_energy_above_hull_per_atom`, `frac_stable_structures`,\n`frac_novel_unique_stable_structures`), and relaxation RMSD when original and\nfinal structures are both available. `energy_above_hull` is per atom; stability\nuses the implementation's 0.1 eV\u002Fatom threshold. Detailed output exposes\nper-structure columns such as `energy_above_hull`,\n`self_consistent_energy_above_hull`, `stable`, `novel`, `unique`, and RMSD.\n\nMatterSim energies are MLFF predictions, not DFT energies. Agreement with DFT\nis not guaranteed, especially for uncommon chemistry; confirm important claims\nwith an explicitly documented DFT workflow. A TRI2024 result is not directly\ncomparable with MP2020 metrics: corrections alter entries and therefore hulls,\nstability, and any stability-conditioned novelty\u002Funiqueness aggregate.\n\n## Recovery rules\n\n- Import or dependency failures: inspect [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md), then run the wrapper's `--help` and a small import\u002Fdevice probe before an expensive job.\n- Optional backend failures: use CPU only for a small diagnostic; relaxation is\n  substantially more expensive and may require CUDA plus compatible MatterSim.\n- Input\u002Fconfig failures: fix suffixes, lowercase directory extensions, ZIP\n  contents, energy shape\u002Forder, reference `.gz` path, and mode conflicts first.\n- Matcher failures: use `disordered` for any partial occupancy or disordered\n  structure; an ordered matcher is not valid when the evaluator detects a\n  disordered structure.\n- Reference\u002Fcorrection failures: rebuild or select a reference with the same\n  correction convention; do not silence compatibility assertions.\n- Workflow failures: preserve input files and partial metrics, rerun a small\n  subset or no-relax path, and only then retry relaxation. Do not make the\n  wrapper fetch a missing reference or potential implicitly.\n\n## Linked runtime material\n\n- [`scripts\u002Fevaluate_structures.py`](scripts\u002Fevaluate_structures.py) — safe,\n  lazy-importing CLI adapter.\n- [`references\u002Fapi-reference.md`](references\u002Fapi-reference.md) — API, defaults,\n  flags, and metric contract.\n- [`references\u002Fworkflows.md`](references\u002Fworkflows.md) — evaluation recipes,\n  output precedence, and resource boundaries.\n- [`references\u002Fdata-formats.md`](references\u002Fdata-formats.md) — accepted input,\n  energy, reference, and output serialization.\n- [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md) — actionable\n  diagnosis and recovery.\n","# MatterGen evaluation\n\nUse this sub-skill when the task is to score generated crystal structures for\nvalidity, uniqueness, novelty, stability, energy above hull, relaxation RMSD,\nor related metrics. It covers MatterGen 1.0.3's evaluation API and CLI. It\ndoes not generate structures, run DFT, or acquire large model\u002Freference assets.\n\n## Start with the mode decision\n\nChoose exactly one mode before loading data:\n\n- **Relaxed evaluation** (`relax=True`): pass structures to MatterSim's\n  `BatchRelaxer`; its final structures and total energies are evaluated. Do not\n  pass `energies`\u002F`--energies_path` in this mode.\n- **Precomputed-energy evaluation** (`relax=False`): pass already relaxed\n  structures and one externally computed total energy per structure. The array\n  must have the same count and order as the loader result.\n\nThe implementation raises if both `relax=True` and energies are supplied, even\nthough an older docstring says energies would be ignored. Use the safe wrapper\nat [`scripts\u002Fevaluate_structures.py`](scripts\u002Fevaluate_structures.py) for\ninput\u002Fmode validation and clear errors; it only invokes package APIs when the\nscript is explicitly run.\n\nRead the applicable reference before running:\n\n- [`references\u002Fworkflows.md`](references\u002Fworkflows.md) for recipes, output\n  precedence, and expensive\u002Fnetwork boundaries.\n- [`references\u002Fapi-reference.md`](references\u002Fapi-reference.md) for exact\n  Python and CLI parameters, defaults, and metrics.\n- [`references\u002Fdata-formats.md`](references\u002Fdata-formats.md) for structure,\n  energy-array, reference, and output formats.\n- [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md) for\n  install\u002Fbackend, data, matcher, correction, and workflow failures.\n\n## Minimal preflight\n\n1. Confirm the input path exists and is one of `.xyz`, `.extxyz`, `.zip`, or a\n   directory containing lowercase `.cif`, `.xyz`, or `.extxyz` files.\n2. Record the loader order. A direct XYZ\u002FEXTXYZ file is read frame-by-frame;\n   a ZIP is extracted temporarily; a directory is traversed in the exact\n   `os.listdir` order (not sorted). Keep any energy array in that same order.\n3. For `relax=False`, load a one-dimensional, numeric, finite `energies.npy`\n   and verify `len(energies) == len(structures)` before calling `evaluate`.\n4. Select a matcher: `disordered` is the robust default for generated alloys,\n   partial occupancies, and mixed ordered\u002Fdisordered sets; use `ordered` only\n   when the data and reference are genuinely ordered and strict reduced-formula\n   grouping is intended.\n5. Select a correction\u002Freference pair. MP2020 is the historical benchmark\n   pairing. TRI2024 (`TRI110Compatibility2024`) plus the TRI2024 reference is\n   recommended for work outside historical benchmarking. Never mix a\n   correction scheme with a reference corrected under another scheme.\n6. Check device, MatterSim availability\u002Fcache, disk space for outputs, and\n   whether a large relaxation is affordable. The wrapper does not download\n   datasets or potential files.\n\n## Canonical API\n\nThe package function is:\n\n```python\nevaluate(\n    structures, relax=True, energies=None, reference=None,\n    structure_matcher=DefaultDisorderedStructureMatcher(),\n    save_as=None, save_detailed_as=None, potential_load_path=None,\n    device=\"cuda\", structures_output_path=None,\n    energy_correction_scheme=MaterialsProject2020Compatibility(),\n)\n```\n\nIt returns a dictionary of aggregate metric values. With `reference=None`, the\nmetric evaluator selects the MP2020 preset, or the TRI2024 preset when the\nTRI2024 correction object is supplied. A supplied reference is deserialized by\nthe CLI from the gzip-compressed LMDB format. Missing terminal systems can make\nenergy metrics unavailable while structure metrics remain computable; treat\nwarnings and absent metrics as results to investigate, not as zeros.\n\n## CLI routing\n\nThe safe script uses these exact underscore flags:\n\n```text\n--structures_path PATH                 required\n--relax {True,False}                   default True\n--energies_path PATH                   required in no-relax mode\n--structure_matcher {ordered,disordered}\n--save_as PATH                         aggregate JSON\n--save_detailed_as PATH                per-structure serialized table\n--potential_load_path PATH             local MatterSim checkpoint\n--reference_dataset_path PATH          .gz LMDB reference\n--device DEVICE                        auto, cpu, cuda[:N], or mps\n--structures_output_path PATH          relaxed EXTXYZ, relax mode only\n--energy_correction_scheme {MP2020,TRI2024}\n```\n\nExample relaxed evaluation:\n\n```bash\npython \u003Cmattergen-skill-root>\u002Fsub-skills\u002Fevaluation\u002Fscripts\u002Fevaluate_structures.py \\\n  --structures_path results \\\n  --relax True --structure_matcher disordered \\\n  --reference_dataset_path data\u002Freference_TRI2024correction.gz \\\n  --energy_correction_scheme TRI2024 \\\n  --save_as results\u002Fmetrics.json \\\n  --save_detailed_as results\u002Fdetailed_metrics.json \\\n  --structures_output_path results\u002Frelaxed.extxyz\n```\n\nExample with external energies:\n\n```bash\npython \u003Cmattergen-skill-root>\u002Fsub-skills\u002Fevaluation\u002Fscripts\u002Fevaluate_structures.py \\\n  --structures_path results.extxyz --relax False \\\n  --energies_path energies.npy --structure_matcher disordered \\\n  --reference_dataset_path data\u002Freference_TRI2024correction.gz \\\n  --energy_correction_scheme TRI2024 --save_as metrics.json\n```\n\n## Interpret the result\n\nThe evaluator can report structure metrics (`precision`, `recall`,\n`frac_unique_structures`, `frac_novel_structures`, `frac_novel_unique_structures`,\nchemical-system novelty, validity, and composition validity), energy metrics\n(`avg_energy_above_hull_per_atom`, `frac_stable_structures`,\n`frac_novel_unique_stable_structures`), and relaxation RMSD when original and\nfinal structures are both available. `energy_above_hull` is per atom; stability\nuses the implementation's 0.1 eV\u002Fatom threshold. Detailed output exposes\nper-structure columns such as `energy_above_hull`,\n`self_consistent_energy_above_hull`, `stable`, `novel`, `unique`, and RMSD.\n\nMatterSim energies are MLFF predictions, not DFT energies. Agreement with DFT\nis not guaranteed, especially for uncommon chemistry; confirm important claims\nwith an explicitly documented DFT workflow. A TRI2024 result is not directly\ncomparable with MP2020 metrics: corrections alter entries and therefore hulls,\nstability, and any stability-conditioned novelty\u002Funiqueness aggregate.\n\n## Recovery rules\n\n- Import or dependency failures: inspect [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md), then run the wrapper's `--help` and a small import\u002Fdevice probe before an expensive job.\n- Optional backend failures: use CPU only for a small diagnostic; relaxation is\n  substantially more expensive and may require CUDA plus compatible MatterSim.\n- Input\u002Fconfig failures: fix suffixes, lowercase directory extensions, ZIP\n  contents, energy shape\u002Forder, reference `.gz` path, and mode conflicts first.\n- Matcher failures: use `disordered` for any partial occupancy or disordered\n  structure; an ordered matcher is not valid when the evaluator detects a\n  disordered structure.\n- Reference\u002Fcorrection failures: rebuild or select a reference with the same\n  correction convention; do not silence compatibility assertions.\n- Workflow failures: preserve input files and partial metrics, rerun a small\n  subset or no-relax path, and only then retry relaxation. Do not make the\n  wrapper fetch a missing reference or potential implicitly.\n\n## Linked runtime material\n\n- [`scripts\u002Fevaluate_structures.py`](scripts\u002Fevaluate_structures.py) — safe,\n  lazy-importing CLI adapter.\n- [`references\u002Fapi-reference.md`](references\u002Fapi-reference.md) — API, defaults,\n  flags, and metric contract.\n- [`references\u002Fworkflows.md`](references\u002Fworkflows.md) — evaluation recipes,\n  output precedence, and resource boundaries.\n- [`references\u002Fdata-formats.md`](references\u002Fdata-formats.md) — accepted input,\n  energy, reference, and output serialization.\n- [`references\u002Ftroubleshooting.md`](references\u002Ftroubleshooting.md) — actionable\n  diagnosis and recovery.\n",{"name":5,"license":14,"metadata":23,"description":7,"disable-model-invocation":25},{"disco-role":24},"operating",true,"main","ac3fe1afa80fb9a09775ecfb2b6cc3ba850a2db6","skills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation",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 evaluation",{"tool":39,"command":40},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill evaluation",{"tool":42,"command":43},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill evaluation","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002FSKILL.md",[46,52,56,60,64],{"path":47,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":50,"sourceHtmlUrl":51},"references\u002Fapi-reference.md","REFERENCE","参考文档",5226,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002Freferences\u002Fapi-reference.md",{"path":53,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":54,"sourceHtmlUrl":55},"references\u002Fdata-formats.md",4301,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002Freferences\u002Fdata-formats.md",{"path":57,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":58,"sourceHtmlUrl":59},"references\u002Ftroubleshooting.md",5688,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002Freferences\u002Ftroubleshooting.md",{"path":61,"resourceType":48,"resourceTypeDesc":49,"sizeBytes":62,"sourceHtmlUrl":63},"references\u002Fworkflows.md",4678,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002Freferences\u002Fworkflows.md",{"path":65,"resourceType":66,"resourceTypeDesc":67,"sizeBytes":68,"sourceHtmlUrl":69},"scripts\u002Fevaluate_structures.py","SCRIPT","脚本",10935,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fmattergen\u002Fsub-skills\u002Fevaluation\u002Fscripts\u002Fevaluate_structures.py",[71,79,86,93,100,107],{"skillNo":72,"name":73,"nameZh":6,"description":74,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"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":80,"name":81,"nameZh":6,"description":82,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":83,"categories":84,"keywords":85,"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":87,"name":88,"nameZh":6,"description":89,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":90,"categories":91,"keywords":92,"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":94,"name":95,"nameZh":6,"description":96,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":97,"categories":98,"keywords":99,"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":101,"name":102,"nameZh":6,"description":103,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":104,"categories":105,"keywords":106,"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":108,"name":109,"nameZh":6,"description":110,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":75,"licenseFileUrl":6,"sourceHtmlUrl":111,"categories":112,"keywords":113,"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",[],[]]