Review & Update Tag Taxonomy

Run this prompt every ~6 months to keep tags.json aligned with what you actually write about.

When to run

• After accumulating 30+ new entries since last review
• When you notice entries getting tagged with wrong/irrelevant tags
• When a new topic area emerges that doesn't fit existing tags

Prompt

Copy everything below the line into Claude (or any LLM). Attach til.md and tags.json as context.

---

You are a taxonomy curator for a personal TIL (Things I Learned) knowledge base.

Input:

• til.md — all dated entries, each ending with #tag1 #tag2 etc.
• tags.json — current tag definitions with keywords and regex patterns

Your job: Analyze usage patterns and produce an updated tags.json. Follow these rules strictly.

Analysis steps

1. Tag frequency. Count how many entries use each tag. Flag:

   • Dead tags (0-1 uses) — candidate for removal or merge
   • Overloaded tags (>25% of entries) — candidate for splitting

2. Untagged or single-tagged entries. List entries with 0 or 1 tags. Suggest tags or flag if no existing tag fits (signals a missing tag).

3. Cluster analysis. Look for recurring topics that don't have a tag yet. Evidence: 3+ entries sharing a keyword/theme that isn't captured by any current tag.

4. Keyword audit. For each tag, check:

   • Are there keywords that cause false positives? (keyword appears in entries that shouldn't have this tag)
   • Are there entries correctly tagged that don't match any keyword? (missing keyword)
   • Do any regex patterns over-match or under-match?

5. Tag overlap. Identify tag pairs that always co-occur (>80%). Consider merging them.

Output format

Produce a structured report with these sections:

## Tag Health Report

### Remove or merge
- `#tag` → reason, suggested merge target

### Split
- `#tag` → suggested children, reason

### New tags to add
- `#tag` — description, evidence (list 3+ entries that need it)

### Keyword changes
- `#tag`: add keywords [...], remove keywords [...], add patterns [...], reason

### Updated tags.json
(full updated JSON)

Constraints

• Keep total tags between 15-25. Fewer is better.
• Every tag must have ≥3 entries using it. No aspirational tags.
• Prefer merging over splitting. Split only when a tag covers genuinely unrelated topics.
• Don't create tags for technologies you used once — those belong as keywords under a broader tag.
• Preserve the excludeUrls mechanism for citation links that cause false positives.
• Keywords must be word-boundary safe (no partial matches like "db" matching "Adobe").
• Test each keyword mentally: "Would this word appear in an entry that should NOT have this tag?" If yes, make it a pattern with context instead.

Reference: tag format in tags.json

{
  "tag-name": {
    "description": "One line explaining the scope",
    "keywords": ["exact", "word", "matches"],
    "excludeUrls": ["domain.com"],
    "patterns": ["regex patterns for contextual matches"]
  }
}