Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Show recent entries from the LLM Wiki log (wiki/log.md). Uses the standardized
.claude/skills/alirezarezvani-wiki-log/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 237% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 101% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -23% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -32% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -24% | 0% |
<!-- canonical copy: engineering/llm-wiki/commands/wiki-log.md — keep in sync -->
Show recent entries from wiki/log.md. Every LLM operation on the wiki leaves a standardized entry:
## [YYYY-MM-DD] <op> | <title>
<optional detail>/wiki-log # last 10 entries
/wiki-log --last 20
/wiki-log --op ingest --last 10 # only ingest entries
/wiki-log --op lint # recent lint passes
/wiki-log --since 2026-04-01Parses wiki/log.md and prints matching entries. No LLM involvement needed — this is essentially:
bashgrep "^## \[" wiki/log.md | tail -N
…plus optional filters for op type and date range.
ingest — a source was read and integratedquery — a question was answered (when filed back)lint — a health check rancreate — a new page was created outside an ingestupdate — a page was updated outside an ingestdelete — a page was removednote — freeform note (contradictions flagged, thesis revisions, etc.)## [2026-04-11] lint | weekly health check
3 contradictions, 12 orphans, 2 broken links. Fixed broken links; left contradictions for next session.
## [2026-04-10] ingest | Anthropic Monosemanticity
Added sources/monosemanticity.md. Updated concepts/sparse-autoencoder, concepts/polysemanticity, entities/anthropic.
## [2026-04-09] query | SAE vs probing
Filed back to comparisons/sae-vs-probing.md.grep + tail directly on wiki/log.md. No dedicated script needed; that's the point of the standardized header format.→ engineering/llm-wiki/skills/llm-wiki/SKILL.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 | 3,883 | 1,679 | -57% | 1 | 1 | 0% | 624 | 730 | +17% | 0 | 0 | — |
case-02 | fail→pass | 3,478 | 5,523 | +59% | 1 | 1 | 0% | 449 | 1,513 | +237% | 0 | 0 | — |
case-03 | fail→fail | 5,824 | 3,862 | -34% | 1 | 1 | 0% | 601 | 854 | +42% | 0 | 0 | — |
case-04 | fail→pass | 2,327 | 1,255 | -46% | 1 | 1 | 0% | 349 | 700 | +101% | 0 | 0 | — |
case-05 | fail→pass | 10,306 | 4,937 | -52% | 1 | 1 | 0% | 1,673 | 1,286 | -23% | 0 | 0 | — |
case-06 | fail→pass | 9,154 | 3,514 | -62% | 1 | 1 | 0% | 1,591 | 1,076 | -32% | 0 | 0 | — |
case-07 | fail→pass | 8,216 | 2,798 | -66% | 1 | 1 | 0% | 1,500 | 1,137 | -24% | 0 | 0 | — |
case-08 | fail→pass | 8,903 | 1,509 | -83% | 1 | 1 | 0% | 1,608 | 766 | -52% | 0 | 0 | — |
case-09 | fail→pass | 11,581 | 3,671 | -68% | 1 | 1 | 0% | 1,883 | 1,073 | -43% | 0 | 0 | — |
case-10 | pass→pass | 8,068 | 2,845 | -65% | 1 | 1 | 0% | 1,346 | 988 | -27% | 0 | 0 | — |
case-11 | fail→pass | 8,086 | 2,759 | -66% | 1 | 1 | 0% | 1,415 | 969 | -32% | 0 | 0 | — |
case-12 | fail→pass | 8,030 | 3,179 | -60% | 1 | 1 | 0% | 1,567 | 1,088 | -31% | 0 | 0 | — |
case-13 | pass→pass | 5,721 | 2,022 | -65% | 1 | 1 | 0% | 1,187 | 836 | -30% | 0 | 0 | — |
case-14 | fail→pass | 11,792 | 2,372 | -80% | 1 | 1 | 0% | 1,838 | 887 | -52% | 0 | 0 | — |
case-15 | pass→pass | 8,783 | 2,668 | -70% | 1 | 1 | 0% | 1,351 | 960 | -29% | 0 | 0 | — |
case-16 | fail→pass | 5,493 | 1,651 | -70% | 1 | 1 | 0% | 1,010 | 762 | -25% | 0 | 0 | — |
case-17 | fail→pass | 12,301 | 2,269 | -82% | 1 | 1 | 0% | 2,161 | 960 | -56% | 0 | 0 | — |
case-18 | pass→pass | 6,471 | 3,225 | -50% | 1 | 1 | 0% | 1,073 | 1,130 | +5% | 0 | 0 | — |
case-19 | fail→pass | 7,500 | 1,485 | -80% | 1 | 1 | 0% | 1,249 | 783 | -37% | 0 | 0 | — |
case-20 | fail→pass | 7,276 | 3,922 | -46% | 1 | 1 | 0% | 1,194 | 1,241 | +4% | 0 | 0 | — |
case-21 | fail→pass | 7,842 | 3,814 | -51% | 1 | 1 | 0% | 1,385 | 1,213 | -12% | 0 | 0 | — |
case-22 | fail→pass | 9,674 | 2,673 | -72% | 1 | 1 | 0% | 1,598 | 1,041 | -35% | 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. The headline lift of +73 percentage points is the difference between those two pass rates over the 22 comparable cases.
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.