Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Assembles a multi-tier context bundle for a given topic. Searches across all nodes, collects relevant files, and loads them at appropriate tiers (L0 = 100 tokens, L1 = 2K tokens, L2 = full). Returns a structured bundle with token counts so you know exactly what you're loading.
.claude/skills/miosa-osa-assemble/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -7% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -60% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -77% | 0% |
| case-06 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-08 | ✗→✓ | ▲ Improved | -36% | 0% |
Build a tiered context bundle for a topic.
Searches for relevant contexts, loads L0+L1+L2 with token counts, assembles into a coherent context bundle optimized for the AI's context window.
/assemble "AI Masters pricing"
/assemble "platform architecture" --limit 10| Flag | Description | Default | |------|-------------|---------| | --limit | Max contexts to include | 10 |
bashcd engine && mix optimal.assemble "topic"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,570 | 17,072 | +10% | 1 | 1 | 0% | 2,618 | 2,437 | -7% | 0 | 0 | — |
case-02 | fail→fail | 17,680 | 6,563 | -63% | 1 | 1 | 0% | 2,434 | 431 | -82% | 0 | 0 | — |
case-03 | fail→pass | 17,680 | 4,487 | -75% | 1 | 1 | 0% | 2,295 | 918 | -60% | 0 | 0 | — |
case-04 | pass→pass | 24,764 | 1,999 | -92% | 1 | 1 | 0% | 4,701 | 513 | -89% | 0 | 0 | — |
case-05 | fail→pass | 15,773 | 1,620 | -90% | 1 | 1 | 0% | 1,792 | 408 | -77% | 0 | 0 | — |
case-06 | fail→pass | 11,850 | 2,245 | -81% | 1 | 1 | 0% | 1,637 | 562 | -66% | 0 | 0 | — |
case-07 | pass→pass | 17,385 | 12,345 | -29% | 1 | 1 | 0% | 2,665 | 2,090 | -22% | 0 | 0 | — |
case-08 | fail→pass | 27,488 | 7,010 | -74% | 1 | 1 | 0% | 2,081 | 1,324 | -36% | 0 | 0 | — |
case-09 | fail→pass | 28,222 | 2,642 | -91% | 1 | 1 | 0% | 2,518 | 574 | -77% | 0 | 0 | — |
case-10 | fail→pass | 9,288 | 4,182 | -55% | 1 | 1 | 0% | 1,404 | 719 | -49% | 0 | 0 | — |
case-11 | fail→pass | 10,733 | 3,754 | -65% | 1 | 1 | 0% | 1,573 | 643 | -59% | 0 | 0 | — |
case-12 | fail→pass | 8,754 | 3,371 | -61% | 1 | 1 | 0% | 1,400 | 624 | -55% | 0 | 0 | — |
case-13 | fail→pass | 51,414 | 4,368 | -92% | 1 | 1 | 0% | 1,599 | 819 | -49% | 0 | 0 | — |
case-14 | fail→pass | 17,703 | 2,105 | -88% | 1 | 1 | 0% | 2,224 | 387 | -83% | 0 | 0 | — |
case-15 | pass→pass | 18,867 | 2,825 | -85% | 1 | 1 | 0% | 1,343 | 529 | -61% | 0 | 0 | — |
case-16 | pass→pass | 11,288 | 5,881 | -48% | 1 | 1 | 0% | 1,449 | 818 | -44% | 0 | 0 | — |
case-17 | fail→pass | 30,870 | 2,511 | -92% | 1 | 1 | 0% | 5,012 | 458 | -91% | 0 | 0 | — |
case-18 | pass→pass | 12,008 | 3,788 | -68% | 1 | 1 | 0% | 1,649 | 650 | -61% | 0 | 0 | — |
case-19 | fail→fail | 9,748 | 10,907 | +12% | 1 | 1 | 0% | 1,299 | 1,967 | +51% | 0 | 0 | — |
case-20 | fail→pass | 12,249 | 8,761 | -28% | 1 | 1 | 0% | 1,620 | 1,395 | -14% | 0 | 0 | — |
case-21 | fail→pass | 7,996 | 4,036 | -50% | 1 | 1 | 0% | 1,202 | 701 | -42% | 0 | 0 | — |
case-22 | pass→pass | 12,546 | 3,718 | -70% | 1 | 1 | 0% | 2,008 | 741 | -63% | 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 +64 percentage points is the difference between those two pass rates over the 21 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.