Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Query the LLM Wiki — reads index.md first, drills into 3-10 relevant pages, synthesizes an answer with inline [[wikilink]] citations, and offers to file the answer back as a new comparison or synthesis page. Usage /wiki-query "<question>"
.claude/skills/alirezarezvani-wiki-query/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-14 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-18 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -22% | 0% |
| case-22 | ✗→✓ | ▲ Improved | -18% | 0% |
| case-04 | ✓→✗ | ▼ Worse | -55% | 0% |
<!-- canonical copy: engineering/llm-wiki/commands/wiki-query.md — keep in sync (root copy uses repo-root-relative script paths) -->
Ask the wiki a question. The librarian reads index.md first, picks relevant pages across categories, synthesizes an answer with citations, and offers to file the answer back into the wiki so your explorations compound.
/wiki-query "<your question>"
/wiki-query "what does the wiki say about sparse autoencoders?"
/wiki-query "compare monosemanticity and polysemanticity across my sources"
/wiki-query "which sources disagree on scaling laws?"
/wiki-query "give me a comparison table of SAE vs linear probing"wiki/index.md to find relevant pagesengineering/llm-wiki/skills/llm-wiki/scripts/wiki_search.py (BM25)[[sources/xxx]] citations + "Related pages" sectioncomparisons/ or synthesis/)The answer's format follows the question:
| Question shape | Output | |---|---| | "What is X?" | Markdown explanation with citations | | "A vs B" | Comparison table | | "Give me a slide deck on X" | Markdown synthesis → /wiki-marp to render | | "Chart the trend in X" | Python script + saved chart in wiki/assets/charts/ |
This command dispatches the wiki-librarian sub-agent. See agents/wiki-librarian.md.
engineering/llm-wiki/skills/llm-wiki/scripts/wiki_search.py — BM25 fallback searchengineering/llm-wiki/skills/llm-wiki/scripts/append_log.py — log filed answers[[wikilink]].→ engineering/llm-wiki/skills/llm-wiki/SKILL.md → engineering/llm-wiki/skills/llm-wiki/references/query-workflow.md
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | pass→fail | 11,272 | 1,929 | -83% | 1 | 1 | 0% | 2,028 | 917 | -55% | 0 | 0 | — |
case-01 | fail→fail | 11,623 | 4,478 | -61% | 1 | 1 | 0% | 1,850 | 871 | -53% | 0 | 0 | — |
case-02 | fail→fail | 2,094 | 5,551 | +165% | 1 | 1 | 0% | 356 | 980 | +175% | 0 | 0 | — |
case-03 | fail→fail | 14,099 | 6,148 | -56% | 1 | 1 | 0% | 2,870 | 1,075 | -63% | 0 | 0 | — |
case-05 | pass→fail | 3,690 | 7,676 | +108% | 1 | 1 | 0% | 731 | 1,025 | +40% | 0 | 0 | — |
case-06 | pass→fail | 11,204 | 6,343 | -43% | 1 | 1 | 0% | 2,036 | 1,139 | -44% | 0 | 0 | — |
case-07 | fail→fail | 22,993 | 5,806 | -75% | 1 | 1 | 0% | 3,955 | 1,052 | -73% | 0 | 0 | — |
case-08 | fail→fail | 10,971 | 5,124 | -53% | 1 | 1 | 0% | 2,103 | 900 | -57% | 0 | 0 | — |
case-09 | fail→fail | 19,701 | 2,717 | -86% | 1 | 1 | 0% | 3,553 | 998 | -72% | 0 | 0 | — |
case-10 | fail→fail | 12,278 | 6,124 | -50% | 1 | 1 | 0% | 2,314 | 855 | -63% | 0 | 0 | — |
case-11 | fail→fail | 9,856 | 4,444 | -55% | 1 | 1 | 0% | 1,658 | 827 | -50% | 0 | 0 | — |
case-12 | pass→pass | 3,729 | 2,709 | -27% | 1 | 1 | 0% | 659 | 1,107 | +68% | 0 | 0 | — |
case-13 | fail→fail | 17,467 | 5,445 | -69% | 1 | 1 | 0% | 2,807 | 913 | -67% | 0 | 0 | — |
case-14 | fail→pass | 5,094 | 1,635 | -68% | 1 | 1 | 0% | 837 | 895 | +7% | 0 | 0 | — |
case-15 | fail→fail | 11,546 | 5,056 | -56% | 1 | 1 | 0% | 2,040 | 909 | -55% | 0 | 0 | — |
case-16 | fail→fail | 9,743 | 5,009 | -49% | 1 | 1 | 0% | 1,499 | 1,030 | -31% | 0 | 0 | — |
case-17 | fail→fail | 14,999 | 5,840 | -61% | 1 | 1 | 0% | 1,941 | 994 | -49% | 0 | 0 | — |
case-18 | fail→pass | 7,351 | 1,798 | -76% | 1 | 1 | 0% | 1,201 | 988 | -18% | 0 | 0 | — |
case-19 | fail→pass | 8,223 | 2,202 | -73% | 1 | 1 | 0% | 1,263 | 982 | -22% | 0 | 0 | — |
case-20 | fail→fail | 13,315 | 19,668 | +48% | 1 | 1 | 0% | 2,205 | 1,018 | -54% | 0 | 0 | — |
case-21 | fail→fail | 10,311 | 4,595 | -55% | 1 | 1 | 0% | 2,162 | 849 | -61% | 0 | 0 | — |
case-22 | fail→pass | 13,022 | 6,320 | -51% | 1 | 1 | 0% | 2,192 | 1,792 | -18% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 8 counted toward the lift figure. The other 14 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +5 percentage points is the difference between those two pass rates over the 8 comparable cases. 10 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.