Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Clear the knowledge base and start fresh
.claude/skills/aiskillstore-learn-reset/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 109% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -38% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -26% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 322% | 0% |
Clear all accumulated knowledge and reset to a fresh state.
knowledge/learnings/ files (patterns, quirks, decisions)Warning: This action cannot be undone. All accumulated insights will be lost.
knowledge/learnings/patterns.mdknowledge/learnings/quirks.mdknowledge/learnings/decisions.mdknowledge/cache/classifications.mdknowledge/context/session.mdknowledge/state.json to initial valuesAfter reset, each learnings file should have:
yaml--- type: [type] version: "1.0" description: [original description] last_updated: null entry_count: 0 --- # [Title] [Description] **Purpose:** [Purpose] --- <!-- Entries will be appended below this line -->
Reset knowledge/state.json to:
json{ "version": "1.0", "learning_mode": false, "learning_mode_since": null, "last_extraction": null, "extraction_count": 0, "queries_since_extraction": 0, "extraction_threshold_queries": 10, "extraction_threshold_minutes": 30 }
Knowledge Base Reset
────────────────────
Are you sure you want to clear all knowledge? This cannot be undone.
[After confirmation]
Knowledge base has been reset:
- Cleared 8 patterns
- Cleared 3 quirks
- Cleared 5 decisions
- Cleared 23 cached classifications
- Reset learning state
The knowledge base is now empty. Use /learn to start fresh.| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-06 | pass→pass | 5,749 | 2,348 | -59% | 1 | 1 | 0% | 906 | 997 | +10% | 0 | 0 | — |
case-01 | fail→pass | 4,370 | 4,566 | +4% | 1 | 1 | 0% | 617 | 1,287 | +109% | 0 | 0 | — |
case-02 | fail→pass | 6,876 | 2,064 | -70% | 1 | 1 | 0% | 987 | 871 | -12% | 0 | 0 | — |
case-03 | fail→fail | 5,314 | 2,709 | -49% | 1 | 1 | 0% | 776 | 996 | +28% | 0 | 0 | — |
case-04 | fail→pass | 10,687 | 3,122 | -71% | 1 | 1 | 0% | 1,829 | 1,143 | -38% | 0 | 0 | — |
case-05 | fail→pass | 8,704 | 2,631 | -70% | 1 | 1 | 0% | 1,379 | 1,020 | -26% | 0 | 0 | — |
case-07 | pass→pass | 9,713 | 1,815 | -81% | 1 | 1 | 0% | 1,503 | 828 | -45% | 0 | 0 | — |
case-08 | fail→pass | 2,493 | 2,716 | +9% | 1 | 1 | 0% | 234 | 988 | +322% | 0 | 0 | — |
case-09 | pass→pass | 2,733 | 1,336 | -51% | 1 | 1 | 0% | 371 | 713 | +92% | 0 | 0 | — |
case-10 | fail→pass | 6,803 | 2,464 | -64% | 1 | 1 | 0% | 1,025 | 880 | -14% | 0 | 0 | — |
case-11 | fail→pass | 8,339 | 1,746 | -79% | 1 | 1 | 0% | 1,273 | 832 | -35% | 0 | 0 | — |
case-12 | fail→pass | 11,457 | 4,147 | -64% | 1 | 1 | 0% | 1,677 | 1,181 | -30% | 0 | 0 | — |
case-13 | fail→fail | 10,405 | 1,359 | -87% | 1 | 1 | 0% | 1,550 | 721 | -53% | 0 | 0 | — |
case-14 | fail→fail | 10,086 | 4,401 | -56% | 1 | 1 | 0% | 1,663 | 946 | -43% | 0 | 0 | — |
case-15 | pass→pass | 7,473 | 2,493 | -67% | 1 | 1 | 0% | 1,085 | 886 | -18% | 0 | 0 | — |
case-16 | pass→pass | 12,360 | 2,947 | -76% | 1 | 1 | 0% | 1,718 | 980 | -43% | 0 | 0 | — |
case-17 | pass→pass | 9,666 | 3,229 | -67% | 1 | 1 | 0% | 1,416 | 954 | -33% | 0 | 0 | — |
case-18 | pass→pass | 4,492 | 1,744 | -61% | 1 | 1 | 0% | 568 | 795 | +40% | 0 | 0 | — |
case-19 | fail→pass | 7,690 | 2,889 | -62% | 1 | 1 | 0% | 1,119 | 996 | -11% | 0 | 0 | — |
case-20 | pass→pass | 9,725 | 5,213 | -46% | 1 | 1 | 0% | 1,481 | 1,346 | -9% | 0 | 0 | — |
case-21 | pass→fail | 11,414 | 6,380 | -44% | 1 | 1 | 0% | 1,699 | 708 | -58% | 0 | 0 | — |
case-22 | pass→pass | 3,929 | 6,607 | +68% | 1 | 1 | 0% | 432 | 1,663 | +285% | 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 21 counted toward the lift figure. The other 1 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 +36 percentage points is the difference between those two pass rates over the 21 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.