Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Ingest a source file from raw/ into the LLM Wiki — read, discuss, write summary page, update cross-references across 5-15 pages, regenerate index, append to log. Usage /wiki-ingest <path-to-source>
.claude/skills/alirezarezvani-wiki-ingest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | -28% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -33% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -30% | 0% |
<!-- canonical copy: engineering/llm-wiki/commands/wiki-ingest.md — keep in sync (root copy uses repo-root-relative script paths) -->
Ingest a new source into the LLM Wiki. This is the most-used command.
The flow: read the source → discuss TL;DR and key claims with you → write a source summary page → update every relevant entity and concept page → flag contradictions → update index.md → append to log.md.
A typical ingest touches 5-15 wiki pages. You (the user) are in the loop: the ingestor proposes changes and waits for your confirmation before writing.
/wiki-ingest <path>
/wiki-ingest raw/papers/monosemanticity.pdf
/wiki-ingest raw/articles/2026-04-01-interpretability-post.mdengineering/llm-wiki/skills/llm-wiki/scripts/ingest_source.py to get title, preview, and suggested summary pathengineering/llm-wiki/skills/llm-wiki/scripts/update_index.py or edits wiki/index.md inlineengineering/llm-wiki/skills/llm-wiki/scripts/append_log.py --op ingest --title "<title>"This command dispatches the wiki-ingestor sub-agent for the heavy lifting. See agents/wiki-ingestor.md.
engineering/llm-wiki/skills/llm-wiki/scripts/ingest_source.py — source prep (metadata + preview)engineering/llm-wiki/skills/llm-wiki/scripts/update_index.py — regenerate indexengineering/llm-wiki/skills/llm-wiki/scripts/append_log.py — log the ingestraw/ layer. If it isn't, the command will ask you to move it first.raw/ is immutable — the ingestor reads only.→ engineering/llm-wiki/skills/llm-wiki/SKILL.md → engineering/llm-wiki/skills/llm-wiki/references/ingest-workflow.md
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 20,124 | 6,211 | -69% | 1 | 1 | 0% | 3,233 | 1,008 | -69% | 0 | 0 | — |
case-02 | fail→fail | 4,870 | 4,648 | -5% | 1 | 1 | 0% | 359 | 960 | +167% | 0 | 0 | — |
case-03 | fail→fail | 5,776 | 6,907 | +20% | 1 | 1 | 0% | 306 | 1,111 | +263% | 0 | 0 | — |
case-04 | fail→pass | 10,561 | 4,218 | -60% | 1 | 1 | 0% | 2,133 | 1,538 | -28% | 0 | 0 | — |
case-05 | pass→pass | 7,621 | 5,910 | -22% | 1 | 1 | 0% | 1,437 | 1,927 | +34% | 0 | 0 | — |
case-06 | pass→pass | 10,455 | 8,985 | -14% | 1 | 1 | 0% | 1,820 | 2,255 | +24% | 0 | 0 | — |
case-07 | fail→pass | 10,226 | 2,913 | -72% | 1 | 1 | 0% | 1,785 | 1,203 | -33% | 0 | 0 | — |
case-08 | pass→pass | 4,808 | 2,895 | -40% | 1 | 1 | 0% | 838 | 1,233 | +47% | 0 | 0 | — |
case-09 | fail→pass | 6,719 | 2,524 | -62% | 1 | 1 | 0% | 1,099 | 1,125 | +2% | 0 | 0 | — |
case-10 | fail→pass | 6,355 | 1,927 | -70% | 1 | 1 | 0% | 1,023 | 1,010 | -1% | 0 | 0 | — |
case-11 | fail→pass | 11,015 | 2,529 | -77% | 1 | 1 | 0% | 1,616 | 1,124 | -30% | 0 | 0 | — |
case-12 | fail→pass | 10,367 | 1,677 | -84% | 1 | 1 | 0% | 1,850 | 916 | -50% | 0 | 0 | — |
case-13 | pass→pass | 7,375 | 3,125 | -58% | 1 | 1 | 0% | 1,200 | 1,139 | -5% | 0 | 0 | — |
case-14 | fail→pass | 11,424 | 1,677 | -85% | 1 | 1 | 0% | 1,882 | 968 | -49% | 0 | 0 | — |
case-15 | fail→fail | 7,112 | 2,068 | -71% | 1 | 1 | 0% | 1,178 | 994 | -16% | 0 | 0 | — |
case-16 | pass→pass | 10,940 | 7,986 | -27% | 1 | 1 | 0% | 1,708 | 2,088 | +22% | 0 | 0 | — |
case-17 | pass→pass | 5,612 | 3,398 | -39% | 1 | 1 | 0% | 964 | 1,215 | +26% | 0 | 0 | — |
case-18 | pass→pass | 6,396 | 2,258 | -65% | 1 | 1 | 0% | 1,205 | 1,049 | -13% | 0 | 0 | — |
case-19 | fail→fail | 7,287 | 1,522 | -79% | 1 | 1 | 0% | 1,235 | 950 | -23% | 0 | 0 | — |
case-20 | fail→pass | 7,426 | 1,676 | -77% | 1 | 1 | 0% | 1,177 | 962 | -18% | 0 | 0 | — |
case-21 | fail→pass | 7,522 | 4,005 | -47% | 1 | 1 | 0% | 1,254 | 1,351 | +8% | 0 | 0 | — |
case-22 | pass→pass | 11,322 | 4,781 | -58% | 1 | 1 | 0% | 2,126 | 1,534 | -28% | 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 19 counted toward the lift figure. The other 3 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 +41 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is 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.