[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M797JSAPPPHYSQBA9XDW0P8V":3,"skill-same:01M797JSAPPPHYSQBA9XDW0P8V":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},"01M797JSAPPPHYSQBA9XDW0P8V","classifier-visualizers",null,"Use Yellowbrick classification diagnostics for scikit-learn classifiers, probability curves, class labels, class balance, pipelines, and saved visual reports.","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\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: classifier-visualizers\ndescription: \"Use Yellowbrick classification diagnostics for scikit-learn\n  classifiers, probability curves, class labels, class balance, pipelines, and\n  saved visual reports.\"\ndisable-model-invocation: true\nmetadata:\n  disco-role: operating\nlicense: Apache 2.0\n---\n\n# Yellowbrick Classifier Visualizers\n\n## Use this sub-skill when\n\n- You need visual diagnostics for a scikit-learn **classifier** rather than a\n  regressor, clusterer, feature matrix, or text corpus.\n- The user asks for a classification report, confusion matrix, ROC-AUC curve,\n  precision-recall curve, class prediction error chart, discrimination threshold\n  curve, class balance chart, or classifier quick method.\n- The workflow involves train\u002Ftest splits, fitted estimators, class labels,\n  label encoders, probability or decision scores, pipelines, or saved PNG\u002FSVG\n  classifier figures.\n\n## Route elsewhere when\n\n- The request is about display setup, `Agg`, axes reuse, style palettes,\n  notebooks, or generic `show(outpath=...)` behavior: first read\n  [visualizer patterns](..\u002F..\u002Freferences\u002Fvisualizer-patterns.md), then use this\n  sub-skill for classifier-specific choices.\n- The failure is a package import, backend, font, Matplotlib display, or broad\n  scikit-learn compatibility issue: read root\n  [troubleshooting](..\u002F..\u002Freferences\u002Ftroubleshooting.md) before applying the\n  classifier-specific fixes here.\n- The task is target-only EDA beyond classifier support, such as broader target\n  distributions or feature-target correlations: route to\n  [feature-target visualizers](..\u002Ffeature-target-visualizers\u002FSKILL.md).\n- The user wants experimental decision-boundary plots or contrib wrappers:\n  route to [contrib and extensions](..\u002Fcontrib-and-extensions\u002FSKILL.md), not to\n  core `yellowbrick.classifier` score visualizers.\n\n## Classifier visualizer map\n\n| Need | Use | Important constraints |\n|---|---|---|\n| Per-class precision\u002Frecall\u002FF1 heatmap | `ClassificationReport` | Needs classifier `predict`; `support` controls support column. |\n| True-vs-predicted matrix | `ConfusionMatrix` | Use `percent=True` for row percentages; avoid percent with filtered classes. |\n| ROC\u002FAUC curves | `ROCAUC` \u002F `roc_auc` | Needs `predict_proba` or `decision_function`; configure binary vs multiclass. |\n| Precision-recall curves | `PrecisionRecallCurve` \u002F `PRCurve` | Needs `predict_proba` or `decision_function`; use for imbalanced classes. |\n| Wrong-class distribution | `ClassPredictionError` | Binary\u002Fmulticlass only; class filtering is not fully supported. |\n| Threshold tuning | `DiscriminationThreshold` | Binary classifiers only; repeats split\u002Ffit trials on clones. |\n| Target support before\u002Fafter split | `ClassBalance` \u002F `class_balance` | Target-only, no estimator; use `labels` matching discovered classes. |\n\n## Required read order for future agents\n\n1. Read [API reference](references\u002Fapi-reference.md) to choose the correct\n   visualizer, import path, signature, and parameter settings.\n2. Read [workflows](references\u002Fworkflows.md) for end-to-end recipes covering\n   reports, binary\u002Fmulticlass ROC and PR, threshold tuning, class balance,\n   pipelines, and headless saving.\n3. If anything fails, read classifier-specific\n   [troubleshooting](references\u002Ftroubleshooting.md), then root\n   [troubleshooting](..\u002F..\u002Freferences\u002Ftroubleshooting.md) for shared display or\n   dependency issues.\n4. For a safe local smoke check, run\n   [classification_smoke.py](scripts\u002Fclassification_smoke.py). It uses synthetic\n   data, forces Matplotlib `Agg`, performs no network access, and writes PNGs.\n\n## Canonical lifecycle\n\nUse the explicit visualizer lifecycle when you need control over fitting,\nscoring, saving, or already-fitted estimators:\n\n```python\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.linear_model import LogisticRegression\nfrom yellowbrick.classifier import ClassificationReport\n\nX_train, X_test, y_train, y_test = train_test_split(\n    X, y, test_size=0.25, stratify=y, random_state=42\n)\n\nmodel = LogisticRegression(max_iter=1000)\nviz = ClassificationReport(model, classes=class_names, support=True)\nviz.fit(X_train, y_train)      # fits the wrapped estimator unless already fitted\nscore = viz.score(X_test, y_test)  # draws and returns estimator score\u002Fmetric\nviz.show(outpath=\"classification_report.png\", clear_figure=True)\n```\n\nPrefer the explicit lifecycle over quick methods when saving files, when the\nestimator is already fitted, when you need to reuse axes, or when you need to\ncatch scores and visualizer attributes such as `scores_`, `confusion_matrix_`,\n`roc_auc`, `score_`, `predictions_`, or `support_`.\n\n## Quick-method policy\n\nYellowbrick quick methods instantiate the visualizer, call `fit`, call `score`\nwhen applicable, optionally call `show()`, and return the visualizer:\n\n- `classification_report(...)`\n- `confusion_matrix(...)`\n- `roc_auc(...)`\n- `precision_recall_curve(...)`\n- `class_prediction_error(...)`\n- `discrimination_threshold(...)`\n- `class_balance(...)`\n\nUse quick methods for notebook-style one-liners or simple scripts. Pass\n`show=False` when embedding in a pipeline step or when you will save later with\n`viz.show(outpath=..., clear_figure=True)`. For score visualizer quick methods,\nprovide both `X_test` and `y_test`, or neither; passing only one raises a\nYellowbrick value error.\n\n## Binary vs multiclass decisions\n\n- `ClassificationReport`, `ConfusionMatrix`, and `ClassPredictionError` support\n  ordinary binary and multiclass classifier targets.\n- `ROCAUC` supports binary and multiclass single-output targets. For a binary\n  estimator with one-dimensional `decision_function` scores, set `binary=True`\n  for one positive-class curve, or set `micro=False, macro=False` and choose\n  `per_class` deliberately.\n- Multiclass `ROCAUC` requires at least one of `micro`, `macro`, or `per_class`\n  to be true; defaults draw per-class plus micro\u002Fmacro curves.\n- `PrecisionRecallCurve` draws a binary curve for binary targets. For multiclass\n  targets it wraps the estimator with a one-vs-rest strategy; use\n  `per_class=True, micro=False` for individual class curves, or keep the default\n  `micro=True, per_class=False` for the micro-average curve.\n- `DiscriminationThreshold` is binary only. It rejects multiclass targets and\n  non-probability\u002Fnon-decision-score estimators.\n- `ClassBalance` is target-only and works before any model is created.\n\n## Class labels and encoders\n\nUse `classes` only to provide display names in the order Yellowbrick discovers\nsorted classes. Use `encoder` when the target values are encoded and you need a\nmapping from encoded values to human-readable labels. If both are supplied,\nYellowbrick prefers the encoder and may warn.\n\nFor `ConfusionMatrix`, `ClassificationReport`, `ROCAUC`, `PrecisionRecallCurve`,\nand `ClassPredictionError`, fit on representative labels so that train\u002Ftest\nclasses align with the estimator's `classes_`. If a rare class is missing from a\nsplit, use a stratified split or provide more data rather than forcing labels.\nFor target-only `ClassBalance`, `labels` must have the same length as the unique\nclasses discovered across `y_train` and `y_test`.\n\n## Probability and decision-score requirements\n\n`ROCAUC`, `PrecisionRecallCurve`, and `DiscriminationThreshold` need score-like\nestimator outputs:\n\n- Good defaults: `LogisticRegression`, `GaussianNB`, `BernoulliNB`, many tree\n  ensembles with `predict_proba`.\n- `SVC` needs `probability=True` for `predict_proba`; otherwise use its\n  `decision_function` where compatible.\n- `LinearSVC` and `RidgeClassifier` can work for some ROC\u002FPR cases through\n  `decision_function`, but binary curve flags may need adjustment.\n- Plain classifiers with only `predict` cannot produce ROC, PR, or threshold\n  curves; use a calibrated wrapper or a model exposing scores.\n\n## Pipeline use\n\nTwo patterns are safe:\n\n1. Wrap the complete preprocessing-plus-classifier `Pipeline` as the estimator:\n   `ClassificationReport(pipe, classes=names).fit(X_train, y_train)`.\n2. Put a visualizer as the final pipeline step when you want a scikit-learn\n   pipeline whose final `.score()` draws the figure. Keep visualizers at the end;\n   they are not feature transformers for downstream model steps.\n\nFor quick methods, pass the complete pipeline as `estimator`. Keep train\u002Ftest\narrays in the same feature representation expected by the pipeline's first step.\n\n## Saving and headless execution\n\n- Use `matplotlib.use(\"Agg\")` before importing `pyplot` in scripts and CI.\n- Prefer `viz.show(outpath=\"plot.png\", clear_figure=True)` for saved files.\n- Reuse explicit Matplotlib axes only when combining plots intentionally; most\n  classifier diagnostics are easiest as separate figures.\n- Use [classification_smoke.py](scripts\u002Fclassification_smoke.py) to verify that\n  the installed package, compatible scikit-learn stack, and `Agg` output path can\n  generate classifier PNGs without external data.\n\n## Validation checklist\n\nBefore handing a classifier visualization to a user:\n\n- Confirm the estimator is a classifier and, for ROC\u002FPR\u002Fthreshold, exposes\n  `predict_proba` or `decision_function`.\n- Confirm binary-only visualizers are not receiving multiclass targets.\n- Confirm class display names match discovered classes and are stable across\n  train\u002Ftest splits.\n- Confirm the split is stratified when class support matters.\n- Confirm `show(outpath=...)` wrote a non-empty file in headless contexts.\n- If a normal scikit-learn classifier is rejected as \"not a classifier\", check\n  root dependency compatibility guidance before rewriting code.\n","# Yellowbrick Classifier Visualizers\n\n## Use this sub-skill when\n\n- You need visual diagnostics for a scikit-learn **classifier** rather than a\n  regressor, clusterer, feature matrix, or text corpus.\n- The user asks for a classification report, confusion matrix, ROC-AUC curve,\n  precision-recall curve, class prediction error chart, discrimination threshold\n  curve, class balance chart, or classifier quick method.\n- The workflow involves train\u002Ftest splits, fitted estimators, class labels,\n  label encoders, probability or decision scores, pipelines, or saved PNG\u002FSVG\n  classifier figures.\n\n## Route elsewhere when\n\n- The request is about display setup, `Agg`, axes reuse, style palettes,\n  notebooks, or generic `show(outpath=...)` behavior: first read\n  [visualizer patterns](..\u002F..\u002Freferences\u002Fvisualizer-patterns.md), then use this\n  sub-skill for classifier-specific choices.\n- The failure is a package import, backend, font, Matplotlib display, or broad\n  scikit-learn compatibility issue: read root\n  [troubleshooting](..\u002F..\u002Freferences\u002Ftroubleshooting.md) before applying the\n  classifier-specific fixes here.\n- The task is target-only EDA beyond classifier support, such as broader target\n  distributions or feature-target correlations: route to\n  [feature-target visualizers](..\u002Ffeature-target-visualizers\u002FSKILL.md).\n- The user wants experimental decision-boundary plots or contrib wrappers:\n  route to [contrib and extensions](..\u002Fcontrib-and-extensions\u002FSKILL.md), not to\n  core `yellowbrick.classifier` score visualizers.\n\n## Classifier visualizer map\n\n| Need | Use | Important constraints |\n|---|---|---|\n| Per-class precision\u002Frecall\u002FF1 heatmap | `ClassificationReport` | Needs classifier `predict`; `support` controls support column. |\n| True-vs-predicted matrix | `ConfusionMatrix` | Use `percent=True` for row percentages; avoid percent with filtered classes. |\n| ROC\u002FAUC curves | `ROCAUC` \u002F `roc_auc` | Needs `predict_proba` or `decision_function`; configure binary vs multiclass. |\n| Precision-recall curves | `PrecisionRecallCurve` \u002F `PRCurve` | Needs `predict_proba` or `decision_function`; use for imbalanced classes. |\n| Wrong-class distribution | `ClassPredictionError` | Binary\u002Fmulticlass only; class filtering is not fully supported. |\n| Threshold tuning | `DiscriminationThreshold` | Binary classifiers only; repeats split\u002Ffit trials on clones. |\n| Target support before\u002Fafter split | `ClassBalance` \u002F `class_balance` | Target-only, no estimator; use `labels` matching discovered classes. |\n\n## Required read order for future agents\n\n1. Read [API reference](references\u002Fapi-reference.md) to choose the correct\n   visualizer, import path, signature, and parameter settings.\n2. Read [workflows](references\u002Fworkflows.md) for end-to-end recipes covering\n   reports, binary\u002Fmulticlass ROC and PR, threshold tuning, class balance,\n   pipelines, and headless saving.\n3. If anything fails, read classifier-specific\n   [troubleshooting](references\u002Ftroubleshooting.md), then root\n   [troubleshooting](..\u002F..\u002Freferences\u002Ftroubleshooting.md) for shared display or\n   dependency issues.\n4. For a safe local smoke check, run\n   [classification_smoke.py](scripts\u002Fclassification_smoke.py). It uses synthetic\n   data, forces Matplotlib `Agg`, performs no network access, and writes PNGs.\n\n## Canonical lifecycle\n\nUse the explicit visualizer lifecycle when you need control over fitting,\nscoring, saving, or already-fitted estimators:\n\n```python\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.linear_model import LogisticRegression\nfrom yellowbrick.classifier import ClassificationReport\n\nX_train, X_test, y_train, y_test = train_test_split(\n    X, y, test_size=0.25, stratify=y, random_state=42\n)\n\nmodel = LogisticRegression(max_iter=1000)\nviz = ClassificationReport(model, classes=class_names, support=True)\nviz.fit(X_train, y_train)      # fits the wrapped estimator unless already fitted\nscore = viz.score(X_test, y_test)  # draws and returns estimator score\u002Fmetric\nviz.show(outpath=\"classification_report.png\", clear_figure=True)\n```\n\nPrefer the explicit lifecycle over quick methods when saving files, when the\nestimator is already fitted, when you need to reuse axes, or when you need to\ncatch scores and visualizer attributes such as `scores_`, `confusion_matrix_`,\n`roc_auc`, `score_`, `predictions_`, or `support_`.\n\n## Quick-method policy\n\nYellowbrick quick methods instantiate the visualizer, call `fit`, call `score`\nwhen applicable, optionally call `show()`, and return the visualizer:\n\n- `classification_report(...)`\n- `confusion_matrix(...)`\n- `roc_auc(...)`\n- `precision_recall_curve(...)`\n- `class_prediction_error(...)`\n- `discrimination_threshold(...)`\n- `class_balance(...)`\n\nUse quick methods for notebook-style one-liners or simple scripts. Pass\n`show=False` when embedding in a pipeline step or when you will save later with\n`viz.show(outpath=..., clear_figure=True)`. For score visualizer quick methods,\nprovide both `X_test` and `y_test`, or neither; passing only one raises a\nYellowbrick value error.\n\n## Binary vs multiclass decisions\n\n- `ClassificationReport`, `ConfusionMatrix`, and `ClassPredictionError` support\n  ordinary binary and multiclass classifier targets.\n- `ROCAUC` supports binary and multiclass single-output targets. For a binary\n  estimator with one-dimensional `decision_function` scores, set `binary=True`\n  for one positive-class curve, or set `micro=False, macro=False` and choose\n  `per_class` deliberately.\n- Multiclass `ROCAUC` requires at least one of `micro`, `macro`, or `per_class`\n  to be true; defaults draw per-class plus micro\u002Fmacro curves.\n- `PrecisionRecallCurve` draws a binary curve for binary targets. For multiclass\n  targets it wraps the estimator with a one-vs-rest strategy; use\n  `per_class=True, micro=False` for individual class curves, or keep the default\n  `micro=True, per_class=False` for the micro-average curve.\n- `DiscriminationThreshold` is binary only. It rejects multiclass targets and\n  non-probability\u002Fnon-decision-score estimators.\n- `ClassBalance` is target-only and works before any model is created.\n\n## Class labels and encoders\n\nUse `classes` only to provide display names in the order Yellowbrick discovers\nsorted classes. Use `encoder` when the target values are encoded and you need a\nmapping from encoded values to human-readable labels. If both are supplied,\nYellowbrick prefers the encoder and may warn.\n\nFor `ConfusionMatrix`, `ClassificationReport`, `ROCAUC`, `PrecisionRecallCurve`,\nand `ClassPredictionError`, fit on representative labels so that train\u002Ftest\nclasses align with the estimator's `classes_`. If a rare class is missing from a\nsplit, use a stratified split or provide more data rather than forcing labels.\nFor target-only `ClassBalance`, `labels` must have the same length as the unique\nclasses discovered across `y_train` and `y_test`.\n\n## Probability and decision-score requirements\n\n`ROCAUC`, `PrecisionRecallCurve`, and `DiscriminationThreshold` need score-like\nestimator outputs:\n\n- Good defaults: `LogisticRegression`, `GaussianNB`, `BernoulliNB`, many tree\n  ensembles with `predict_proba`.\n- `SVC` needs `probability=True` for `predict_proba`; otherwise use its\n  `decision_function` where compatible.\n- `LinearSVC` and `RidgeClassifier` can work for some ROC\u002FPR cases through\n  `decision_function`, but binary curve flags may need adjustment.\n- Plain classifiers with only `predict` cannot produce ROC, PR, or threshold\n  curves; use a calibrated wrapper or a model exposing scores.\n\n## Pipeline use\n\nTwo patterns are safe:\n\n1. Wrap the complete preprocessing-plus-classifier `Pipeline` as the estimator:\n   `ClassificationReport(pipe, classes=names).fit(X_train, y_train)`.\n2. Put a visualizer as the final pipeline step when you want a scikit-learn\n   pipeline whose final `.score()` draws the figure. Keep visualizers at the end;\n   they are not feature transformers for downstream model steps.\n\nFor quick methods, pass the complete pipeline as `estimator`. Keep train\u002Ftest\narrays in the same feature representation expected by the pipeline's first step.\n\n## Saving and headless execution\n\n- Use `matplotlib.use(\"Agg\")` before importing `pyplot` in scripts and CI.\n- Prefer `viz.show(outpath=\"plot.png\", clear_figure=True)` for saved files.\n- Reuse explicit Matplotlib axes only when combining plots intentionally; most\n  classifier diagnostics are easiest as separate figures.\n- Use [classification_smoke.py](scripts\u002Fclassification_smoke.py) to verify that\n  the installed package, compatible scikit-learn stack, and `Agg` output path can\n  generate classifier PNGs without external data.\n\n## Validation checklist\n\nBefore handing a classifier visualization to a user:\n\n- Confirm the estimator is a classifier and, for ROC\u002FPR\u002Fthreshold, exposes\n  `predict_proba` or `decision_function`.\n- Confirm binary-only visualizers are not receiving multiclass targets.\n- Confirm class display names match discovered classes and are stable across\n  train\u002Ftest splits.\n- Confirm the split is stratified when class support matters.\n- Confirm `show(outpath=...)` wrote a non-empty file in headless contexts.\n- If a normal scikit-learn classifier is rejected as \"not a classifier\", check\n  root dependency compatibility guidance before rewriting code.\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\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers",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 classifier-visualizers",{"tool":40,"command":41},"bunx","bunx skills add VectorSpaceLab\u002FAREX-Skill --skill classifier-visualizers",{"tool":43,"command":44},"pnpm","pnpm dlx skills add VectorSpaceLab\u002FAREX-Skill --skill classifier-visualizers","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers\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",10653,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers\u002Freferences\u002Fapi-reference.md",{"path":54,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":55,"sourceHtmlUrl":56},"references\u002Ftroubleshooting.md",10783,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers\u002Freferences\u002Ftroubleshooting.md",{"path":58,"resourceType":49,"resourceTypeDesc":50,"sizeBytes":59,"sourceHtmlUrl":60},"references\u002Fworkflows.md",10419,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers\u002Freferences\u002Fworkflows.md",{"path":62,"resourceType":63,"resourceTypeDesc":64,"sizeBytes":65,"sourceHtmlUrl":66},"scripts\u002Fclassification_smoke.py","SCRIPT","脚本",7518,"https:\u002F\u002Fgithub.com\u002FVectorSpaceLab\u002FAREX-Skill\u002Fblob\u002Fmain\u002Fskills\u002Frepositories\u002Frepo-skills\u002Fyellowbrick\u002Fsub-skills\u002Fclassifier-visualizers\u002Fscripts\u002Fclassification_smoke.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",[],[]]