القائمة
Use when creating a new skill, editing an existing skill, or verifying a skill works before deployment
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Use when receiving an ambiguous feature request, when scope is unclear, or before writing any plan
Use after verification-before-completion passes and code review is clean, to close out the development branch
Use after writing a plan document in docs/plans/, before requesting user confirmation
Use when receiving feedback from code-reviewer subagent or human reviewer
Use when a feature, bugfix, or refactoring step is completed and needs review, or before merging to main, or when user says "review", "审查", "帮我看看代码"
Use when executing an implementation plan with 3+ independent tasks in the current session
| name | writing-skills |
| description | Use when creating a new skill, editing an existing skill, or verifying a skill works before deployment |
Writing skills IS Test-Driven Development applied to process documentation.
Core principle: If you didn't watch an agent fail without the skill, you don't know if the skill teaches the right thing.
NO SKILL WITHOUT A FAILING TEST FIRST
This applies to NEW skills AND EDITS to existing skills. No exceptions.
AGENTS.md)| Type | Description | Example |
|---|---|---|
| Technique | 具体方法，有步骤可循 | test-driven-development |
| Pattern | 思维模型 | brainstorming |
| Reference | 查询式，结构化条目 | —（本项目暂无 Reference 型 skill） |
| Label | Meaning | Requirement |
|---|---|---|
| Rigid | 必须严格遵守 | 必须带 Rationalizations + Red Flags |
| Flexible | 原则可按场景取舍 | 不需 Rationalizations |
.agents/skills/ ← 唯一存储位置（编辑这里）
<skill-name>/
SKILL.md # 必需
<supporting-file>.md # 仅当 >100 行或为可复用 prompt 模板
.claude/skills → ../.agents/skills ← 软链接，让 Claude Code 默认路径也能读到
只编辑 .agents/skills/ 下的源文件——软链接自动同步。不要在 .claude/skills/ 下新建真实目录（会覆盖软链接）。
SessionStart hook 直接读 .agents/skills/using-growingio-sdk-skills/SKILL.md（见 .claude/hooks/session-start.sh）。
---
name: skill-name-with-hyphens
description: Use when <triggering condition>
---
Description rules:
Use when / Use before / Use afterWHY: When a description summarizes workflow, Claude follows the description instead of reading the full skill body. Description with workflow = agent skips body.
# <Skill Name>
> **Type:** Technique | **Discipline:** Rigid
## Overview
## When to Use / Not to Use
## Process / Checklist
## Rationalizations (Rigid skills required)
| Excuse | Reality |
## Red Flags (Rigid skills required)
## Related Skills
Run pressure scenario WITHOUT skill. Document rationalizations verbatim.
Address those specific rationalizations. Run same scenarios WITH skill.
Agent found new rationalization? Add counter. Re-test.
| Excuse | Reality |
|---|---|
| "Skill is obviously clear, no need to test" | Clear to you ≠ clear to agents |
| "Description needs workflow context" | Workflow in description → agent skips body |
| "Just adding a section, no test needed" | Same Iron Law |
| "I'll test if problems emerge" | Test BEFORE, not after |
.agents/skills/<skill-name>/SKILL.md.claude/skills/<skill-name>/SKILL.md（通过软链接解析）.agents/skills/using-growingio-sdk-skills/SKILL.md（见 .claude/hooks/session-start.sh）这个布局让不同 CLI 工具（各自默认读 .agents/ 或 .claude/）共享同一份文件。不要在 .claude/skills/ 下创建真实目录——会覆盖软链接。