Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Refines an active research session by drilling deeper into a subtopic. Use after tome:research to narrow results to a specific channel or angle.
.claude/skills/athola-dig/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 140% | 0% |
| case-03 | ✗→✓ | ▲ Improved | 468% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -59% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 846% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -51% | 0% |
/tome:research first)Skill(tome:synthesize))Refine an active research session interactively.
/tome:research \"topic\""
suggest available channels
executes; if absent, error "Start a session first with /tome:research \"<topic>\"" is emitted and skill halts
overwriting prior findings from other channels
the updated file reflects the deeper search results
a warning is emitted listing the available channels
saved, not before
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 33,419 | 36,741 | +10% | 1 | 1 | 0% | 6,113 | 738 | -88% | 0 | 0 | — |
case-02 | fail→pass | 11,476 | 25,345 | +121% | 1 | 1 | 0% | 1,917 | 4,606 | +140% | 0 | 0 | — |
case-03 | fail→pass | 3,432 | 18,907 | +451% | 1 | 1 | 0% | 608 | 3,455 | +468% | 0 | 0 | — |
case-04 | fail→pass | 17,616 | 4,273 | -76% | 1 | 1 | 0% | 2,898 | 1,183 | -59% | 0 | 0 | — |
case-05 | fail→fail | 21,556 | 38,664 | +79% | 1 | 1 | 0% | 3,035 | 1,370 | -55% | 0 | 0 | — |
case-06 | fail→pass | 5,573 | 46,202 | +729% | 1 | 1 | 0% | 298 | 2,820 | +846% | 0 | 0 | — |
case-07 | fail→pass | 21,755 | 6,801 | -69% | 1 | 1 | 0% | 3,183 | 1,574 | -51% | 0 | 0 | — |
case-08 | fail→pass | 39,660 | 53,257 | +34% | 1 | 1 | 0% | 7,126 | 3,673 | -48% | 0 | 0 | — |
case-09 | pass→pass | 22,984 | 16,907 | -26% | 1 | 1 | 0% | 3,562 | 2,768 | -22% | 0 | 0 | — |
case-10 | fail→pass | 18,734 | 22,391 | +20% | 1 | 1 | 0% | 3,035 | 3,311 | +9% | 0 | 0 | — |
case-11 | pass→pass | 32,468 | 14,211 | -56% | 1 | 1 | 0% | 4,141 | 2,714 | -34% | 0 | 0 | — |
case-12 | fail→pass | 20,489 | 4,672 | -77% | 1 | 1 | 0% | 3,431 | 924 | -73% | 0 | 0 | — |
case-13 | pass→pass | 27,595 | 22,321 | -19% | 1 | 1 | 0% | 4,819 | 3,475 | -28% | 0 | 0 | — |
case-14 | pass→fail | 22,179 | 7,176 | -68% | 1 | 1 | 0% | 4,260 | 737 | -83% | 0 | 0 | — |
case-15 | fail→fail | 108,788 | 7,576 | -93% | 1 | 1 | 0% | 8,216 | 1,258 | -85% | 0 | 0 | — |
case-16 | fail→pass | 6,121 | 11,383 | +86% | 1 | 1 | 0% | 419 | 2,346 | +460% | 0 | 0 | — |
case-17 | fail→pass | 13,630 | 18,542 | +36% | 1 | 1 | 0% | 1,806 | 3,098 | +72% | 0 | 0 | — |
case-18 | pass→pass | 24,774 | 14,265 | -42% | 1 | 1 | 0% | 3,884 | 2,058 | -47% | 0 | 0 | — |
case-19 | fail→pass | 26,881 | 4,507 | -83% | 1 | 1 | 0% | 3,668 | 981 | -73% | 0 | 0 | — |
case-20 | pass→pass | 45,313 | 19,171 | -58% | 1 | 1 | 0% | 8,224 | 3,033 | -63% | 0 | 0 | — |
case-21 | fail→pass | 14,560 | 15,648 | +7% | 1 | 1 | 0% | 2,267 | 2,262 | -0% | 0 | 0 | — |
case-22 | fail→fail | 31,063 | 20,120 | -35% | 1 | 1 | 0% | 4,199 | 3,205 | -24% | 0 | 0 | — |
case-23 | fail→pass | 19,988 | 26,256 | +31% | 1 | 1 | 0% | 3,641 | 3,618 | -1% | 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. 23 cases were attempted, and 18 counted toward the lift figure. The other 5 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 +52 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.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/22/2026 | +27% |
Other measured skills in the registry, with their headline benchmark lift.