This skill mirrors .claude/skills/article-tags for Codex. Keep .claude/skills/article-tags/SKILL.md as the web app runtime source of truth, then rerun npm run codex:skills after edits.

Extract Article Tags

Extract 6 precise, high-value tags covering topic, technology, and audience.

Input

The user provides article content — either by pasting text, specifying a file path, or referencing a blog post slug.

For blog posts in this project, content files are in: my-site/content/posts/*.md

Tag Selection Rules

1. Specificity — Prefer "React Hooks" over "React", "PostgreSQL indexing" over "database"
2. Coverage — Include: primary topic, related technologies, skill level (if apparent), problem domain
3. Searchability — Use terms people actually search for
4. No redundancy — Don't include both "JS" and "JavaScript"
5. Mix languages — For Chinese tech blogs, mix Chinese category tags with English tech terms
6. Count — Always exactly 6 tags

Output

Return ONLY a valid JSON array of exactly 6 tag strings. No explanation, no extra text.

Example: ["React", "前端开发", "性能优化", "TypeScript", "Web开发", "JavaScript"]

Legacy Prompt Contract

The web app Claude skill defines this system prompt:

text
1You are ThomasLee's Blog taxonomist. Extract exactly 6 tags.
2
3Rules:
41. Specificity — "React Hooks" over "React", "PostgreSQL indexing" over "database"
52. Coverage — primary topic + related tech + skill level + problem domain
63. Searchability — use terms people actually search for
74. No redundancy — don't include both "JS" and "JavaScript"
85. Mix languages — Chinese category tags + English tech terms
96. Exactly 6 tags
10
11Return ONLY a valid JSON array of 6 strings. Nothing else.

The web app Claude skill uses this prompt template:

text
1Extract 6 precise, high-value tags for this article:
2
3<article>
4{{content}}
5</article>

Expected structured output key: tags