Taiwanmd Validate
Testing & Qualitytaiwanmd-validateTaiwanmd Validate is an Agent Skill for testing & quality published in the frank890417/taiwan-md repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Validate Taiwan.md articles for editorial quality, frontmatter correctness, and content standards. Use when reviewing PRs that touch knowledge/ or src/content/, writing new articles, or when the user asks to check article quality. Runs frontmatter validation, word count check, and reports issues. TRIGGER when: user says "validate", "check article", "review quality", or when editing .md files in knowledge/ or src/content/.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/frank890417/taiwan-md/blob/HEAD/.claude/skills/taiwanmd-validate/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/taiwanmd-validate, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
taiwanmd-validate: Article Quality Validation
Validate Taiwan.md articles against editorial standards, frontmatter schema, and content quality requirements.
Validation steps
1. Frontmatter validation
cd "$(git rev-parse --show-toplevel)" && node scripts/test-frontmatter.mjs
Checks all articles for:
- Required fields: title, description, date, tags
- Valid date format
- Tags is an array
- No duplicate slugs
- Valid file naming
2. Check a specific article
Read the article and verify manually:
# Read the article
cat "knowledge/<Category>/<filename>.md" | head -30
Check:
-
title— present and descriptive -
description— 1-2 sentences, contains the "counter-intuitive core insight" -
date— valid ISO date -
tags— array with 3-8 relevant tags -
featured— boolean (optional) -
lastHumanReview— set totrueif human-reviewed -
lastVerified— date of last fact-check
3. Content quality check (article-health prose-health)
cd "$(git rev-parse --show-toplevel)" && python3 scripts/tools/article-health.py "knowledge/<Category>/<filename>.md" --check=prose-health
Scores articles 0-10 for AI-generated "hollow" content. Score > 4 = needs rewrite.
Checks for:
- Bullet list padding (lazy structure)
- Missing specific dates/years
- Hollow modifiers ("非常", "極其", "值得注意的是")
- Excessive em-dashes (> 4 per article = AI signal)
- Textbook-style openings
- Missing source references
4. Reference check
cd "$(git rev-parse --show-toplevel)" && node scripts/check-references.mjs
Verifies articles have "參考資料" or "延伸閱讀" sections with URLs.
5. Wikilink check
cd "$(git rev-parse --show-toplevel)" && node scripts/test-wikilinks.mjs
Validates all [[wikilink]] cross-references resolve to existing articles.
Quality standards (EDITORIAL v4)
Every article should:
- Open with a named person — not an institution or abstract concept
- Have a counter-intuitive core insight in the description
- Include specific dates and numbers — not vague references
- Cite sources — 參考資料 section with URLs
- Word count ≥ 1,500 for full articles
- Use wikilinks
[[Article Name]]for cross-references - Multi-perspective for sensitive topics (politics, identity)
When to use
- After writing or editing an article
- When reviewing a PR that modifies
knowledge/files - When the user asks "is this article good enough?"
- Before marking
lastHumanReview: true