GITHUB AGENT SKILLS / 技能收录
给你的 AI 助手
装上专业技能
我们从 GitHub 持续收录、审核并分类优质的 Agent Skills，按职业与使用场景整理，复制一行命令即可安装到 Claude Code、Cursor 等智能体。
按职业浏览 Agent Skills
为不同岗位挑选的高频技能
按使用场景浏览 Agent Skills
从你要完成的任务出发
按创作者浏览 Agent Skills
按 GitHub Stars 排序的优秀创作者
Microsoft
@microsoft
openclaw
@openclaw
OpenAI
@openai
The Apache Software Foundation
@apache
GitHub
@github
Anthropic
@anthropics
AutoGPT
@Significant-Gravitas
Garry Tan
@garrytan全部 Skills 196852
tabular-ensemble
Combines the search's top models into a stacking ensemble. Only when the top genomes are within about 1% of each other. Check: metric cv_score with a minimum.
tabular-eda
Explores the cleaned table and writes insights for the feature step. Usually right after tabular-clean. Check: expects reports/eda.md.
tabular-cluster
Groups rows when there is no target: evolves KMeans/DBSCAN for the best silhouette and exports the code. Only when the data has no target. Check: expects data/processed/clusters.csv.
tabular-clean
Writes src/clean.py: drops audited columns, fixes impossible values, imputes and handles outliers. Usually right after tabular-audit. Check: run src/clean.py, expects data/processed/clean.csv.
tabular-audit
First step for a table: profile it, fix the target and problem type, find leaking and impossible columns. Usually the first step of a tabular plan. Check: expects reports/audit.md.
python-step
Generic plan step: write a standalone Python script under src/ that turns the step's inputs into its outputs, run it, fix it until it succeeds, then call finish_step.
leakage-checklist
How to recognise data leakage: features unavailable at prediction time or proxies for the target.
evolve-search
Searches models and hyperparameters with the genetic algorithm (or Optuna for expensive models). Usually right after tabular-features. Check: metric mu_cv with a minimum.
evaluate-explain
Holds out data, checks for leakage (health) and warns on overfitting and dominance, and explains the model. Usually the last step of a supervised plan. Check: metric health, min 1.
drift-check
Compares a new batch with the data the model was built on. Only when the inputs include a reference set and a newer batch. Check: expects reports/drift.md.
class-imbalance
What to do when one class is rare.
choose-metric
How to pick the evaluation metric for a tabular problem.
test-skill
Basic local test skill for Skillfile validation
skeuomorphic-forge
This skill should be used when the user asks to build physically-realistic skeuomorphic UI with Tailwind CSS — buttons, panels, gauges, knobs, CRT/LED displays, glass/metal effects, particle systems, industrial hardware. Triggers on requests mentioning: skeuomorphic, realistic depth, industrial UI, 3D button, gauge, meter, analog, tactile, material texture, retro-industrial, aerospace panel, DSP cockpit, VU meter, rotary knob, CRT display, rim light, metal recess, faceplate. Mandatory workflow: resolve bundled resources, read golden examples, run/search references, inspect a matching HTML asset when available, classify the physical object and shadow tier, then verify U0-U7 before delivery. Do NOT trigger for flat/minimal UI or standard Material/Shadcn components.
skeuomorphic-forge
暂无简介
release-notes
Generate release notes from commit messages since the last beta tag. Use when preparing notes for an upcoming beta or production release.
feature-to-requirements
Runs the full feature pipeline - discovers the feature concept with the user, saves it to a file, then runs requirements elicitation based on that concept.
commit
Commit changes with a concise and meaningful message