claude

ml

ML pipeline skills: feature-change (training/serving schema contract + leakage check), model-eval (multi-season leave-one-out regression gate), ship-model (snapshot, registry, artifact publish, cache flush, rollback). Pairs with the ml-reviewer subagent.

6 on GitHubby Engineeringv1.0.0MITupdated yesterday
/

Plugin ml from marketplace personal in kulykivska/claude-plugins. Commands read from the repo's manifest — yesterday.

What it installs

Skills (3)

  • feature-change

    Add, remove or rename a model feature without breaking serving. Trigger on "add a model feature", "new feature in the model", "add a feature", "remove this feature", or any diff touching the feature schema. Covers the training and serving sides of the contract, the leakage check, and deploy ordering

  • model-eval

    Prove a model change is real before shipping it. Trigger on "did it get better?", "check the model", "evaluate this model change", "did this improve MAE", or before any PR that claims a metric win. Runs the multi-season leave-one-out harness and applies the regression gate.

  • ship-model

    Release a retrained model into serving. Trigger on "ship this model", "publish the model", "publish the new weights", or after a model change has passed the eval gate. Covers snapshot, registry, artifact publish, verification against live, and the rollback path.

View source on GitHub

More from kulykivska/claude-plugins

All 21 plugins →

sdlc

claude

Task lifecycle: new-task (BA authoring + contradiction gate + Gherkin), requirements, plan-task, task-review, debug, qa skills + architect, debugger and requirements-analyst subagents.

8 skills · 3 agents

6

content

claude

content-writer and fact-checker subagents + social-post, linkedin-post-writer, humanizer and remove-ai-marks skills: one idea into platform-native posts for LinkedIn, Threads, X, Instagram; EN/UK, owner's voice, fact-checked, humanized, and stripped of AI provenance marks.

4 skills · 2 agents

6

product-design

claude

Selling and retention design: conversion-review, paywall-and-pricing, habit-loop, landing-builder and ux-copy skills + conversion-reviewer subagent. Honest by default (no dark patterns), every finding names the event that proves it.

5 skills · 1 agent

6

reviewers

claude

Per-stack reviewer subagents: Python/FastAPI, web (Vite/React/TS), SwiftUI, and ML pipeline (LORO gate, FEATURE_COLS sync, leakage).

4 agents

6

ux-craft

claude

Apple-grade UX and the toy feel: delight-spec (spring, haptic and motion tokens with CSS and SwiftUI snippets), hig-review, and the design-first redesign orchestrator (5-6 directions, design system, screens, then plan) + hig-reviewer subagent.

3 skills · 1 agent

6

biz

claude

Business growth subagents: aso-optimizer (App Store), growth-analyst (funnel/conversion), outreach-writer (partnership and consulting pitches).

3 agents

6

Maintain ml?

This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.

claude & codex plugins21 plugins