Minimal Patch with Subagents

Minimal patching workflow that debates fix options across parallel independent subagents and applies the smallest safe diff. Use only when you want risk-free code, and to code real minimal that only needs to fix the issue.

9 stars 1 forksAI, LLMs & Agents Plugin marketplace

Install

Claude Code plugin
/plugin marketplace add kargnas/skills

The repo is a Claude Code plugin marketplace. Run /plugin afterwards to pick the plugin that includes this skill.

skills CLI (Claude Code, Cursor, Codex and more)
npx skills add kargnas/skills --skill minimal-patch-with-subagents
Manual install
git clone --depth 1 https://github.com/kargnas/skills.git /tmp/skills
mkdir -p ~/.claude/skills
cp -r /tmp/skills/skills/minimal-patch-with-subagents ~/.claude/skills/minimal-patch-with-subagents

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Minimal Patch with Subagents

Minimal patch workflow for any git repository. Goal: fix the confirmed problem with the fewest necessary changes while preserving existing behavior and keeping review risk low. Fix options are debated across 3-5 parallel independent subagents that vote on the least-risky patch location before any edit.

## Core Principle

This is a **minimal-scope** fix. It may be temporary or permanent, but every decision optimizes for:
1. **Fewest logic files in `git status`** (ideal: 1, max: 2) — **PRIMARY METRIC**
2. **Smallest effective diff** = `logic_lines − 0.3 × (comment_lines + md_lines)`, floor 0
3. **Lowest regression risk** for nearby behavior
4. **Fast, focused verification** proportional to the change

**The single source of truth is `git diff` on uncommitted changes.** Every edit decision MUST be justified against the current git diff output. If a file appears in `git status` that shouldn't, it's a scope violation.

**Comments + Markdown docs have NEGATIVE weight — more is strictly better.** Code comments and `.md` file edits (README, AGENTS.md, CHANGELOG, debate logs, etc.) don't just escape the budget — they *actively shrink* the effective patch size. Formula:

> `effective_lines = logic_lines − 0.3 × (comment_lines + md_lines)`  (floor at 0)
Read the full skill on GitHub

Details

Skill name
minimal-patch-with-subagents
Repository
kargnas/skills
Path
skills/minimal-patch-with-subagents/SKILL.md
Repo updated
18 November 2026

FAQ

What does the Minimal Patch with Subagents skill do?

Minimal patching workflow that debates fix options across parallel independent subagents and applies the smallest safe diff. Use only when you want risk-free code, and to code real minimal that only needs to fix the issue.

How do I install the Minimal Patch with Subagents skill?

Claude Code plugin: /plugin marketplace add kargnas/skills. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Minimal Patch with Subagents popular?

It lives in kargnas/skills, which has 9 GitHub stars and 1 forks.

Related skills

AI Ready

9

kargnas/skills

Use only when the user asks to make the repo AI-ready or AI-optimized (AGENTS.md migration, debugging/testing setup, docs). Do not run on your own.

AI, LLMs & Agents

Design Frontend Sangrak

9

kargnas/skills

Use when starting frontend work in a project with no existing design system, when building or theming UI with shadcn/ui or Tailwind CSS, or when asked to review/audit web UI. Default design baseline for new screens, plus container-first responsive rules, URL-first routing, scoped async loading, theme/i18n defaults, and applied UX laws.

Frontend & Design

Git Lore

9

kargnas/skills

Captures, queries, and configures decision context in native Git trailers. Use for non-trivial commit messages, constraints, rejected alternatives, directives, test gaps, Lore history queries, or installing Lore rules in project or global agent instructions.

Git & Code Review

Minimax H3 Prompt

5

kuronzzhan-droid/minimax-h3-prompt-skill

按 MiniMax-H3 的两种有效格式撰写视频生成提示词(时间轴风格 / 官方结构化风格),覆盖 T2VA 文生视频、I2VA·FL2VA·L2VA 首尾帧、full-reference 参考生视频 R2V。只要用户想用本地 ComfyUI 的 H3 出视频、要求写或改 H3 提示词、提到 MiniMax H3 / ref2va / fl2va、给参考图或音色让角色说话唱歌、或抱怨 H3 出片"角色不像/台词乱语/声音不对/动作没跟上节拍",都应使用本技能——它带官方规范原文、两种风格的实战范例和本机显存实测约束。

AI, LLMs & Agents

MCP Code Execution

10

Archive228/lab-skills

Cut agent token consumption by calling MCP servers through generated code instead of direct tool invocations: present tools as typed wrapper files on a filesystem, discover them progressively, and keep intermediate data in the execution environment (Anthropic measured 150,000 → 2,000 tokens, a 98.7% saving). Trigger when an agent connects to many MCP servers/tools, when tool definitions eat a large share of the context window before work starts, when workflows pipe large intermediate results (transcripts, spreadsheets) between tools, or when sensitive data should not enter model context.

AI, LLMs & Agents

Monorepo Agents

10

Archive228/lab-skills

Operating procedure for running coding agents on large codebases and multi-million-line monorepos: agentic search over indexing, hierarchical CLAUDE.md layering, noise exclusion via .claudeignore and permissions.deny, LSP for symbol-level navigation, subagents for exploration/editing splits, and enterprise rollout patterns. Use when working in, configuring, or deploying Claude Code (or a similar agent harness) on a repo too large to hold in context — many services, hundreds of top-level folders, or millions of files.

AI, LLMs & Agents

Browse the most popular Claude skills, all ai, llms & agents skills, or write your own with the skills generator.