AgentSkillsCN

git-workflow

Git 提交、PR 以及分支管理规范，对 AI 归因要求“零容忍”。在以下场景中使用：(1) 创建提交；(2) 编写提交信息；(3) 创建拉取请求；(4) 编写 PR 描述；(5) 分支或合并代码。绝对准则——在任何提交、PR 或问题关闭之前，切勿在提交、PR、问题或文档中包含“Claude”“Anthropic”“AI 生成”字样，亦不得使用带有“由…生成”字样的表情符号，或任何 AI 工具的归因信息。

SKILL.md
--- frontmatter
name: git-workflow
description: Git commit, PR, and branching standards with ZERO TOLERANCE for AI attribution. Use when (1) Creating commits, (2) Writing commit messages, (3) Creating pull requests, (4) Writing PR descriptions, (5) Branching or merging code. ABSOLUTE RULE - NEVER include "Claude", "Anthropic", "AI-generated", emojis with "Generated with", or any AI tool attribution in commits, PRs, issues, or documentation.

Git & PR Workflow Standards

Core Principles

  • •Small PRs: ≤300 LOC changed ideally
  • •Commit often: Early, meaningful commits with clear messages
  • •Linear history: Prefer rebase over merge commits
  • •Professional: No AI attribution, no tool branding, clean descriptions

🚨 ABSOLUTE ZERO-TOLERANCE RULE: NO AI ATTRIBUTION 🚨

NEVER, UNDER ANY CIRCUMSTANCES, include ANY of the following in:

  • •Git commit messages
  • •Pull request descriptions
  • •Issue comments
  • •GitHub discussions
  • •Code comments
  • •Documentation
  • •ANY public-facing repository content

❌ STRICTLY FORBIDDEN TEXT - NEVER USE

  • •"Claude"
  • •"Anthropic"
  • •"claude.com"
  • •"Claude Code"
  • •"Claude Desktop"
  • •"Generated with Claude"
  • •"Co-Authored-By: Claude"
  • •"🤖 Generated with Claude Code"
  • •"Co-Authored-By: Claude noreply@anthropic.com"
  • •Any emoji followed by "Generated with" or "Powered by"
  • •Any reference to AI assistants, AI code generation, or automated tools
  • •Any branding, attribution, or links to third-party AI services

✅ CORRECT FORMAT - ALWAYS USE

  • •Clean, professional commit messages
  • •Clear description of what changed
  • •Explanation of why the change was made
  • •No attribution, no branding, no tool references

Examples

❌ FORBIDDEN COMMIT

code
Add multi-dimension vector support

- Support for 384, 768, 1024, and 1536 dimensions
- Update validation logic for new dimensions
- Add comprehensive test coverage

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

✅ CORRECT COMMIT

code
Add multi-dimension vector support

- Support for 384, 768, 1024, and 1536 dimensions
- Update validation logic for new dimensions
- Add comprehensive test coverage

❌ FORBIDDEN PR DESCRIPTION

markdown
## Summary
Implemented new authentication flow with JWT tokens.

## Test Plan
All tests passing.

---
🤖 Powered by Claude

✅ CORRECT PR DESCRIPTION

markdown
## Summary
Implemented new authentication flow with JWT tokens.

## Test Plan
All tests passing.

Enforcement

  • •ZERO TOLERANCE rule with NO EXCEPTIONS
  • •Every commit must be verified before pushing
  • •Every PR must be reviewed for attribution before creating
  • •If attribution is found, it must be removed immediately via force push or amendment
  • •This rule applies to ALL repositories, ALL projects, ALL commits
  • •Violations compromise the professional appearance of our work

Why This Matters

  • •Maintains professional repository appearance
  • •Avoids unwanted third-party attribution
  • •Ensures our work is presented as our own
  • •Prevents confusion about authorship and ownership
  • •Compliance with company branding guidelines

PR Requirements

Every PR must include:

  • •Problem/Context: What issue are we solving and why?
  • •Solution summary: How does this change address the problem?
  • •Test plan: Commands + results proving functionality
  • •Risk/rollback: Potential issues and how to revert if needed
  • •Story link + Type + Estimate: Link to issue/story with type (feature/bug/chore) and points

Reference Files

See references/ai-attribution-enforcement.md for comprehensive forbidden text list and enforcement details.

See references/pr-templates.md for complete PR templates for feature, bug, and chore PRs.

See references/branch-conventions.md for branch naming patterns and examples.