Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Ingest one source document (or a folder of them) into the llmwiki. Use when the user drops a new markdown file, PDF, or URL into the wiki and asks you to process it. The user will typically say "ingest this", "add this to the wiki", "process this file into the wiki", or point at a file under `raw/`.
.claude/skills/pratiyush-llmwiki-ingest/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-23 | ✗→✓ | ▲ Improved | 20% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 533% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 89% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 282% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 49% | 0% |
Takes a source file (or a folder) and turns it into wiki pages following the Karpathy LLM Wiki pattern:
wiki/sources/<slug>.md summary[[wikilinks]]wiki/index.md, wiki/overview.md, wiki/log.md/wiki-ingest slash commandllmwiki-sync skill runs the converter first, then invokes this skill for each new fileFollow the Ingest Workflow from the repo's CLAUDE.md exactly:
wiki/index.md and wiki/overview.md for contextwiki/sources/<slug>.md using the Source Page Formatwiki/index.md — new entry under ## Sourceswiki/overview.md if substantial new infowiki/entities/<TitleCase>.md)wiki/concepts/<TitleCase>.md)[[wikilinks]] under ## Connections## Contradictionswiki/log.md: ## [YYYY-MM-DD] ingest | <title>When the source is under raw/sessions/ (a session transcript converted by the converter):
## Conversation section verbatim — use it as raw material to summarisewiki/entities/<ProjectSlug>.md with a ## Sessions listwiki/concepts/ — anything the user explicitly lockedtools_used is a candidate entityis_subagent: true — link to the parent session rather than creating a new project entityraw/ is immutable. Never modify files there.## Contradictions.## Connections section with at least one [[wikilink]].title, type, tags, sources, last_updated.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-23 | fail→pass | 5,940 | 2,359 | -60% | 1 | 1 | 0% | 887 | 1,060 | +20% | 0 | 0 | — |
case-01 | fail→fail | 4,453 | 58,340 | +1210% | 1 | 1 | 0% | 771 | 1,069 | +39% | 0 | 0 | — |
case-02 | fail→fail | 3,948 | 41,112 | +941% | 1 | 1 | 0% | 354 | 1,224 | +246% | 0 | 0 | — |
case-03 | fail→fail | 7,292 | 6,431 | -12% | 1 | 1 | 0% | 1,247 | 1,056 | -15% | 0 | 0 | — |
case-04 | fail→fail | 4,528 | 16,430 | +263% | 1 | 1 | 0% | 261 | 3,174 | +1116% | 0 | 0 | — |
case-05 | fail→fail | 6,127 | 4,235 | -31% | 1 | 1 | 0% | 582 | 993 | +71% | 0 | 0 | — |
case-06 | pass→pass | 3,965 | 6,860 | +73% | 1 | 1 | 0% | 627 | 1,674 | +167% | 0 | 0 | — |
case-07 | fail→pass | 4,178 | 8,418 | +101% | 1 | 1 | 0% | 326 | 2,063 | +533% | 0 | 0 | — |
case-08 | pass→pass | 16,085 | 4,585 | -71% | 1 | 1 | 0% | 2,173 | 1,550 | -29% | 0 | 0 | — |
case-09 | pass→pass | 6,306 | 7,155 | +13% | 1 | 1 | 0% | 885 | 1,959 | +121% | 0 | 0 | — |
case-10 | fail→fail | 2,920 | 7,498 | +157% | 1 | 1 | 0% | 489 | 1,974 | +304% | 0 | 0 | — |
case-11 | fail→pass | 5,680 | 5,733 | +1% | 1 | 1 | 0% | 928 | 1,757 | +89% | 0 | 0 | — |
case-12 | fail→pass | 4,170 | 11,320 | +171% | 1 | 1 | 0% | 733 | 2,801 | +282% | 0 | 0 | — |
case-13 | fail→pass | 19,123 | 3,756 | -80% | 1 | 1 | 0% | 905 | 1,352 | +49% | 0 | 0 | — |
case-14 | fail→pass | 9,153 | 3,737 | -59% | 1 | 1 | 0% | 1,356 | 1,353 | -0% | 0 | 0 | — |
case-15 | fail→pass | 5,994 | 3,485 | -42% | 1 | 1 | 0% | 1,017 | 1,275 | +25% | 0 | 0 | — |
case-16 | fail→pass | 8,713 | 5,136 | -41% | 1 | 1 | 0% | 1,450 | 1,431 | -1% | 0 | 0 | — |
case-17 | pass→pass | 10,948 | 6,858 | -37% | 1 | 1 | 0% | 1,726 | 1,711 | -1% | 0 | 0 | — |
case-18 | pass→pass | 11,125 | 4,088 | -63% | 1 | 1 | 0% | 1,773 | 1,276 | -28% | 0 | 0 | — |
case-19 | fail→pass | 9,057 | 2,052 | -77% | 1 | 1 | 0% | 1,464 | 1,016 | -31% | 0 | 0 | — |
case-20 | fail→pass | 11,946 | 6,287 | -47% | 1 | 1 | 0% | 2,002 | 1,766 | -12% | 0 | 0 | — |
case-21 | pass→pass | 12,489 | 7,101 | -43% | 1 | 1 | 0% | 1,891 | 1,860 | -2% | 0 | 0 | — |
case-22 | pass→pass | 4,870 | 4,155 | -15% | 1 | 1 | 0% | 773 | 1,312 | +70% | 0 | 0 | — |
case-24 | pass→pass | 4,558 | 3,315 | -27% | 1 | 1 | 0% | 708 | 1,201 | +70% | 0 | 0 | — |
case-25 | pass→pass | 9,806 | 3,349 | -66% | 1 | 1 | 0% | 1,774 | 1,250 | -30% | 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. 25 cases were attempted, and 18 counted toward the lift figure. The other 7 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 +40 percentage points is the difference between those two pass rates over the 18 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.