wiki-lookup

Wikipedia Quick Lookup

Quick lookup skill for the local Wikipedia data lake (~7M articles stored as plain text).

How to Search

Find articles by title

rg -i "<query>" data/index/titles.txt | head -20

Find articles by category

rg -i "<query>" data/index/categories.txt | head -20

Word-boundary search (avoid partial matches)

rg -i -w "<query>" data/index/titles.txt | head -20

Use -w when a short query returns too many irrelevant hits — it matches whole words only.

OR search (synonyms, related terms)

rg -i -e "<term1>" -e "<term2>" data/index/titles.txt | head -20

Search for multiple terms at once. Great for synonyms or when the article title might use different phrasing.

Read a specific article

Find the path first, then read it:

rg -m1 "^<slug>\t" data/index/paths.txt | cut -f3

Then use the Read tool on the returned path.

Preview an article (without reading the whole thing)

head -50 data/articles/<prefix>/<slug>.txt

Scan the opening to decide if the article is worth a full read.

Search across article content

rg -l -i "<query>" data/articles/ | head -20

Returns file paths of matching articles. Narrow with title/category search first — full-text is powerful but slower.

Content snippets (see matches in context)

rg -i -m2 -C1 "<query>" data/articles/<prefix>/<slug>.txt

-m2 limits to first 2 matches, -C1 shows 1 line of context around each. Useful for checking relevance before a full read.

Count matches for relevance ranking

rg -i -c "<query>" data/articles/ | sort -t: -k2 -nr | head -10

Articles with more mentions are likely more relevant.

AND search (both terms must appear)

rg -l -i "<term1>" data/articles/ | xargs rg -l -i "<term2>" | head -10

Fixed-string search (disable regex)

rg -F -i "<literal>" data/index/titles.txt

Use -F when your query contains regex-special characters (dots, parens, brackets).

Worked Example — Simple Lookup

Task: "What does Wikipedia say about circuit breaker patterns?"

1. Search titles: rg -i "circuit.breaker" data/index/titles.txt
2. Find path: rg -m1 "^circuit-breaker\t" data/index/paths.txt | cut -f3
3. Read the article with the Read tool
4. Return the relevant information

Worked Example — Narrowing Down Results

Task: "Find information about ant colony optimization"

1. Search titles: rg -i "ant" data/index/titles.txt | head -20 — too many results (elephant, antenna, etc.)
2. Narrow with word boundary: rg -i -w "ant" data/index/titles.txt | head -20 — better but still noisy
3. Try more specific: rg -i "ant colony" data/index/titles.txt — found it
4. If title search fails, try content: rg -l -i "ant colony optimization" data/articles/ | head -5
5. Preview before reading: head -50 data/articles/an/ant-colony-optimization.txt
6. Full read with the Read tool

Search Strategy

1. Titles first — instant and gives you article slugs directly
2. Preview before full read — head -50 saves time on irrelevant articles
3. Narrow before broadening — start specific, add -w if too many results, broaden with -e synonyms if too few
4. Content search last — powerful but slower; use after title/category search narrows scope

When to Escalate

If the task requires cross-domain connections, non-obvious structural analogues, or synthesising insights across multiple articles from different fields — spawn the wiki-explorer agent instead. That agent is purpose-built for lateral knowledge exploration.

Examples of when to escalate:

• "Find biological analogues to our load balancing problem"
• "What can ecology teach us about system resilience?"
• "Search across domains for patterns similar to consensus algorithms"