Editorial Review
by @liqiongyu
Comprehensive editorial review for Chinese content before publication. Use when the user asks to 'review my article', 'proofread this', 'check for errors', '...
clawhub install editorial-review📖 About This Skill
name: editorial-review description: "Comprehensive editorial review for Chinese content before publication. Use when the user asks to 'review my article', 'proofread this', 'check for errors', 'editorial review', 'pre-publish check', 'compliance review', or any Chinese text that needs proofreading, grammar checking, fact verification, political sensitivity screening, or advertising law compliance before release. Also activate when the user says '审校', '校对', '审核', '检查错误', '发布前检查', '合规审查', or submits Chinese content asking for quality review. Covers articles, news, blog posts, press releases, marketing copy, social media posts, and any public-facing Chinese text. Even if the user only asks for 'a quick check' or 'look this over', use this skill to ensure thorough coverage."
Editorial Review
Systematic editorial review for Chinese content before publication. Modeled after the publishing industry's "Three Reviews, Three Proofs" (三审三校) quality system, adapted for AI-assisted single-pass comprehensive review.
This skill catches errors that embarrass authors and organizations: typos, grammar mistakes, factual inaccuracies, political landmines, and advertising law violations. It produces a structured review report with every issue located, categorized, and accompanied by a concrete fix.
When to Activate
Review Modes
Select mode based on content length, stakes, and user request. Announce the selected mode before starting.
| Mode | Content Length | Layers | Output | Use When | |------|---------------|--------|--------|----------| | Quick | < 500 chars | Layers 1-3 | Inline comments | Social media, short posts, internal messages | | Standard | 500-5,000 chars | Layers 1-6 | Structured report | Blog posts, articles, newsletters | | Deep | > 5,000 chars or high-stakes | All 8 layers | Full report + summary | Press releases, official statements, regulatory filings, news articles |
Default to Standard. Escalate to Deep if the content is public-facing and high-stakes (government, legal, financial, medical). Use Quick only when explicitly requested or for obviously short, low-stakes content.
Platform Compatibility
Map interactive tools to whatever the current environment provides:
| Platform | Question tool | File write |
|----------|--------------|------------|
| Claude Code | AskUserQuestion | Write / Edit |
| Codex | request_user_input | write_file / edit_file |
| Gemini | ask_user | file tools |
| Other / Claude.ai | Numbered options in chat | inline output |
Language: The review report is written in the same language as the content (Chinese). Skill internals and reference files are in English.
The 8-Layer Review Pipeline
Process layers sequentially. Each layer catches a distinct class of errors. Do NOT skip layers — earlier layers catch surface errors that could mask deeper issues.
For each issue found, record: location (paragraph/sentence number or quote), category (layer + sub-type), original text, suggested fix, and severity (Critical / Major / Minor).
Layer 1: Character Accuracy (文字准确性)
The most fundamental layer. Chinese input methods produce homophone and shape-similar errors that spell-checkers miss.
Check for:
Consult references/typography-and-punctuation.md § "Common Character Errors" for the full error list.
Severity: Typos in titles/headlines = Critical. Body text typos = Major. Rare/debatable cases = Minor.
Layer 2: Punctuation and Typography (标点与排版)
Chinese has its own punctuation system (GB/T 15834-2011) with strict rules that differ from English.
Check for:
Consult references/typography-and-punctuation.md § "Punctuation Rules" and § "CJK-Latin Mixing" for full rules.
Severity: Systematic punctuation errors = Major. Isolated cases = Minor.
Layer 3: Grammar and Logic (语法与逻辑)
Chinese grammar errors (病句) fall into six standard categories. Logical fallacies undermine credibility.
Check for the six error types: 1. Improper word order (语序不当): misplaced modifiers, adverbials after predicates 2. Mismatched collocation (搭配不当): subject-predicate, verb-object, modifier mismatches 3. Missing/redundant components (成分残缺或赘余): missing subject after 使/让/通过, redundant modifiers 4. Tangled structure (结构混乱): mid-sentence subject shifts, blended constructions 5. Ambiguous meaning (表意不明): unclear pronoun references, scope ambiguity 6. Illogical statements (不合逻辑): self-contradictions, one-sided/two-sided mismatches, improper classification in parallel lists
Also check for:
Consult references/grammar-and-logic.md for detailed examples and detection patterns.
Severity: Ambiguity or contradiction = Critical. Grammar errors = Major. Style-level issues = Minor.
Layer 4: Factual Accuracy (事实准确性)
Incorrect facts damage credibility and may create legal liability.
Check for:
When web access is available, verify key claims using search tools. When unavailable, flag unverifiable claims rather than silently passing them.
Consult references/fact-checking.md for the verification methodology.
Severity: Factual errors in key claims = Critical. Unverified statistics = Major. Minor date/name details = Minor.
Layer 5: Political and Territorial Compliance (政治与领土合规)
Content published in or about China must observe strict political sensitivities. Violations can result in content removal, fines, or worse.
Key areas:
Consult references/political-compliance.md for the full framework and examples.
Severity: Territorial sovereignty violations = Critical. Sensitive framing = Major. Borderline references = Minor (flag for human review).
Layer 6: Advertising Law Compliance (广告法合规)
Chinese advertising law (广告法) prohibits absolute claims and deceptive language with fines of 200K-1M yuan.
Prohibited categories:
Exceptions exist — terms used in corporate philosophy, internal product comparisons, product specifications, or backed by documented certifications are permitted.
Consult references/advertising-law.md for the complete prohibited word list and exception rules.
Severity: Clear prohibited terms in commercial content = Critical. Borderline terms = Major. Non-commercial content = Minor (advisory).
Layer 7: Consistency and Style (一致性与风格)
Inconsistency signals carelessness and confuses readers.
Check for:
Severity: Terminology inconsistency = Major. Minor format variations = Minor.
Layer 8: Structure and Readability (结构与可读性)
The final layer evaluates whether the content works as a whole.
Check for:
Severity: Misleading title = Critical. Structural gaps = Major. Readability nits = Minor.
Output Format
Review Report Structure
Always produce the report in this structure:
# Editorial Review Report (审校报告)Overview
Content type: [article / news / press release / marketing copy / ...]
Word count: [X characters]
Review mode: [Quick / Standard / Deep]
Review date: [YYYY-MM-DD] Summary
Critical issues: [N]
Major issues: [N]
Minor issues: [N]
Overall assessment: [Ready for publication / Needs revision / Needs major rework] Issues by Layer
Layer 1: Character Accuracy
| # | Location | Original | Suggested Fix | Severity | Explanation |
|---|----------|----------|---------------|----------|-------------|
| 1 | Para 2, line 3 | "松驰" | "松弛" | Major | Homophone error... |Layer 2: Punctuation and Typography
...[Continue for each applicable layer]
Clean Version (Optional)
[If requested, provide the corrected full text with all fixes applied]Reviewer Notes
[Any observations about overall quality, patterns, or recommendations for the author]
Severity Definitions
| Severity | Meaning | Action | |----------|---------|--------| | Critical | Factual error, political violation, legal risk, or meaning-altering mistake | Must fix before publication | | Major | Clear error that harms quality or credibility | Should fix before publication | | Minor | Style preference, debatable choice, or minor inconsistency | Fix if time permits |
Workflow
1. Receive content — Read the full text before making any comments. Understand the content type, target audience, and publication context. 2. Select mode — Announce: "This is a [length] [content type]. I'll run a [mode] review covering Layers [N-M]." 3. Execute layers sequentially — Work through each layer. Record every issue found. 4. Compile report — Organize issues by layer. Count severities. Assess overall readiness. 5. Present report — Share the structured report. If the user wants a corrected version, produce it. 6. Iterate — If the user revises and resubmits, run a focused re-review on previously flagged issues.
Examples
Example 1: Quick review
User: "帮我看看这条朋友圈：新年到了,祝大家在新的一年里万事如意!我们公司是全国最好的AI公司,欢迎合作."
Review: Two issues found.
Example 2: Standard review trigger
User: "审校一下这篇公众号文章" [pastes 2000-char article]
→ Standard mode, Layers 1-6, full structured report.
Example 3: Deep review trigger
User: "这是要发到官网的新闻稿，帮我仔细检查一下，特别是合规方面" [pastes 8000-char press release]
→ Deep mode, all 8 layers, extra attention to Layers 5-6.
Caveats
💡 Examples
Example 1: Quick review
User: "帮我看看这条朋友圈：新年到了,祝大家在新的一年里万事如意!我们公司是全国最好的AI公司,欢迎合作."
Review: Two issues found.
Example 2: Standard review trigger
User: "审校一下这篇公众号文章" [pastes 2000-char article]
→ Standard mode, Layers 1-6, full structured report.
Example 3: Deep review trigger
User: "这是要发到官网的新闻稿，帮我仔细检查一下，特别是合规方面" [pastes 8000-char press release]
→ Deep mode, all 8 layers, extra attention to Layers 5-6.