Token导航 LogoToken导航TokenDH.com
研究检索权限需确认github未标认证来源可访问许可证需确认审计通过

orthogonalize-pr正交化 pr

Agent Skill

orthogonalize-pr 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

250

周安装

10

GitHub Stars

公开资料未说明

下载量

81
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:orthogonalize-pr(正交化 pr)
来源仓库:https://github.com/corygabrielsen/skills
仓库路径:skills/orthogonalize-pr
安装命令:
npx skills add https://github.com/corygabrielsen/skills --skill orthogonalize-pr
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/corygabrielsen/skills --skill orthogonalize-pr

简介

orthogonalize-pr 用于查找、检索和筛选相关信息。

  • 适合在 Codex、Claude、Cursor、Gemini CLI 中根据关键词或任务场景快速定位候选结果。
  • 通过 npx skills add 命令从指定 GitHub 仓库安装并使用。
  • 安装前建议确认权限范围和维护状态,注意是否会触发联网或文件读写操作。
  • 可结合来源仓库和原始 README 进一步核验具体用法和功能边界。

SKILL.md

Orthogonalize PR

Identify independent change sets within a PR and separate them into branches.

Why This Exists

Many PRs bundle unrelated changes together. A "fix retry logic" PR might also add logging, refactor a helper, and update tests for something else. These orthogonal changes:

  • Make review harder (reviewer context-switches between concerns)
  • Make bisect harder (which change caused the regression?)
  • Make reverts harder (can't revert just the problematic part)

By identifying orthogonal components, you can suggest splitting the PR or understand its independent pieces.

Prerequisites

Run /decompose-pr first. Orthogonalize operates on a decomposed branch where each commit represents a logical change.

Phase 1: Analyze Dependencies

Review each commit from the decomposed branch and map dependencies.

# List commits on the decomposed branch
git log --oneline <base>..HEAD

# For each commit, examine what it touches
git show --stat <commit>
git show <commit>

For each commit, identify:

  1. What it introduces — New types, functions, files
  2. What it depends on — Types/functions from other commits in this branch
  3. What depends on it — Later commits that use what this introduces

Build a dependency table:

| Commit | Introduces | Depends On | Depended By |
|--------|------------|------------|-------------|
| 1 | CacheEntry type | - | 2, 3 |
| 2 | Cache struct | 1 | 3 |
| 3 | Cache integration | 1, 2 | - |
| 4 | RateLimiter type | - | 5 |
| 5 | RateLimiter integration | 4 | - |

Phase 2: Identify Orthogonal Sets

Group commits into orthogonal sets — clusters with internal dependencies but no cross-cluster dependencies.

Algorithm:

  1. Build a graph where commits are nodes and dependencies are edges
  2. Find connected components — each component is an orthogonal set
  3. Commits with no dependencies on other branch commits are their own set

From the example above:

  • Set A: Commits 1, 2, 3 (Cache feature)
  • Set B: Commits 4, 5 (RateLimiter feature)

These sets are orthogonal — they don't depend on each other.

Present to user:

## Orthogonal Sets

| Set | Commits | Description | Files |
|-----|---------|-------------|-------|
| A | 1, 2, 3 | Cache feature | cache.ts, index.ts |
| B | 4, 5 | Rate limiting | limiter.ts, index.ts |

Sets A and B are independent. They touch `index.ts` in different places
but don't share types or call each other.

Get user input: They may want to name the sets differently, or may see dependencies you missed.

Phase 3: Create Branches

Create a branch for each orthogonal set.

# Start from the decomposed branch
git checkout <user>/decompose/<pr-branch>

# For each set, create a branch with just those commits
git checkout -b <user>/orthogonalize/<pr-branch>/cache <base>
git cherry-pick <commit-1> <commit-2> <commit-3>

git checkout -b <user>/orthogonalize/<pr-branch>/rate-limiter <base>
git cherry-pick <commit-4> <commit-5>

Verify each branch:

# Should build independently
# (prettier, cargo fmt, black, etc.)
# (npm run build, cargo check, make build, etc.)

If a branch doesn't build, either:

  • A dependency was missed — revisit Phase 2
  • The changes aren't truly orthogonal — merge the sets

Phase 4: Summarize

Present the orthogonal structure:

## Orthogonalization Complete

Original PR: #123 "Add caching and rate limiting"
Decomposed branch: user/decompose/feature-branch
Orthogonal branches:

| Branch | Commits | Description | Builds? |
|--------|---------|-------------|---------|
| user/orthogonalize/feature-branch/cache | 3 | Cache feature | Yes |
| user/orthogonalize/feature-branch/rate-limiter | 2 | Rate limiting | Yes |

**Recommendation:** These could be two separate PRs:
1. "Add response caching" (cache branch)
2. "Add rate limiting" (rate-limiter branch)

Merging in either order would work. No conflicts expected.

When Sets Aren't Worth Splitting

Sometimes orthogonal sets exist but splitting isn't worthwhile:

  • Too small: A 5-line logging change isn't worth its own PR
  • Conceptually unified: Author bundled them intentionally for a reason
  • Review overhead: Two PRs means twice the CI, review cycles, merge conflicts

Note these cases but don't force splits. The goal is to *identify* orthogonality, not mandate separation.

Edge Cases

Shared base refactor: Sometimes commit 1 is a refactor that enables both Set A and Set B. Options:

  • Include it in both branches (duplication, but each stands alone)
  • Make it a third "foundation" set that both depend on
  • Leave sets coupled if the refactor is small

File-level overlap: Two sets may touch the same file in different places. This is fine — they're still orthogonal if changes don't interact. Git can merge them.

Test coupling: Tests might span multiple sets. Options:

  • Include tests with the code they test
  • Create a separate "tests" set if tests are truly independent
  • Accept some test duplication across branches

Anti-patterns

  • Orthogonalizing without decomposing first (you need to understand the changes)
  • Creating too many tiny sets (defeats the purpose)
  • Missing semantic dependencies (two functions that must exist together)
  • Forcing orthogonality when changes are genuinely coupled

Enter orthogonalization mode now. Confirm which decomposed branch to analyze, then begin Phase 1.

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

34.65%
按下载量换算28

Claude

29.8%
按下载量换算24

Cursor

19.58%
按下载量换算16

Gemini CLI

8.32%
按下载量换算7

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

权限需确认

当前来源未能明确判断权限范围,默认进入异常复核队列。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills