A standardized article editing process that solves the problem of editing work being lost due to session truncation.
When to Use
- The user asks to edit, modify, or optimize an existing article
- Multiple changes need to be made to the article
- Editing work such as content deletion, tone adjustment, and structural reorganization
Execution Process
Step 1: Read the target article in full
- Use the Read tool to read the article in full
- Quickly understand the article's topic, structure, and current state
Step 2: List all modifications
- Based on the user's needs, list all items that need to be modified
- For each item, indicate: location (which paragraph), modification type (delete/rewrite/adjust), and specific content
- Wait for the user to confirm the scope before starting
Step 3: Incremental editing + real-time saving
- Save the file after completing each group of related modifications
- After completing every 3-5 modifications, report progress to the user:
进度:已完成 X/Y 项 ✅ 已完成:[列表] ⏳ 待完成:[列表]
Step 4: Summarize after completion
- List a summary of all changes
- Explain the reason for each modification
- If external information is cited, include the source
Key Principles
- Confirm before starting: do not assume the scope of modifications
- Save while editing: save immediately after each batch of modifications to prevent loss from truncation
- Transparent changes: the user clearly knows what was changed and why
- Maintain style: after modification, it must still match Huasheng's writing style (see SHARED-RULES.md)
Relationship with the Review Process
- After editing is complete, three-pass review (/AI味审校) can be triggered if the user needs it
- The Editing Skill focuses on "changing according to requirements," while the Review Skill focuses on "reducing AI flavor"
Last updated: 2026-02-06
Produced by Huashu | AI Native Coder · Indie Developer WeChat public account "Huasu" | 300k+ followers | AI tools and efficiency improvement Representative work: Little Cat Fill Light (AppStore paid chart Top1) · "One Book to Master DeepSeek"