Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Packages the current working context into a structured handoff document that another agent or session can load to continue work without losing state. Captures: current progress, open decisions, blocking issues, relevant files, and next steps.
.claude/skills/miosa-osa-handoff/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 64% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -48% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -31% | 0% |
| case-16 | ✗→✓ | ▲ Improved | -15% | 0% |
> Transfer context between agents or sessions with structured state preservation.
/handoff --to <agent|session> [--context "<summary>"] [--files "<file-list>"]Packages the current working context into a structured handoff document that another agent or session can load to continue work without losing state. Captures: current progress, open decisions, blocking issues, relevant files, and next steps.
work/handoff-<timestamp>.md.bash# Hand off to another agent /handoff --to "backend-go" --context "Auth flow designed, need implementation" # Hand off between sessions /handoff --to session --files "lib/auth.ex, test/auth_test.exs" # Quick handoff with inline context /handoff --to "test-automator" --context "Feature implemented, needs test coverage"
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 6,078 | 7,501 | +23% | 1 | 1 | 0% | 933 | 1,483 | +59% | 0 | 0 | — |
case-02 | fail→fail | 16,794 | 13,821 | -18% | 1 | 1 | 0% | 2,366 | 1,901 | -20% | 0 | 0 | — |
case-03 | fail→fail | 15,300 | 13,392 | -12% | 1 | 1 | 0% | 2,445 | 1,965 | -20% | 0 | 0 | — |
case-04 | pass→pass | 6,517 | 5,771 | -11% | 1 | 1 | 0% | 1,026 | 1,226 | +19% | 0 | 0 | — |
case-05 | pass→pass | 9,409 | 13,416 | +43% | 1 | 1 | 0% | 1,483 | 2,068 | +39% | 0 | 0 | — |
case-06 | pass→pass | 8,978 | 15,910 | +77% | 1 | 1 | 0% | 1,531 | 3,135 | +105% | 0 | 0 | — |
case-07 | fail→pass | 4,432 | 7,593 | +71% | 1 | 1 | 0% | 839 | 1,376 | +64% | 0 | 0 | — |
case-08 | fail→fail | 14,555 | 8,482 | -42% | 1 | 1 | 0% | 1,913 | 1,700 | -11% | 0 | 0 | — |
case-09 | fail→pass | 7,054 | 6,947 | -2% | 1 | 1 | 0% | 1,094 | 1,451 | +33% | 0 | 0 | — |
case-10 | fail→pass | 5,997 | 1,734 | -71% | 1 | 1 | 0% | 1,113 | 574 | -48% | 0 | 0 | — |
case-11 | pass→pass | 14,525 | 5,703 | -61% | 1 | 1 | 0% | 2,575 | 1,025 | -60% | 0 | 0 | — |
case-12 | pass→pass | 16,322 | 6,382 | -61% | 1 | 1 | 0% | 2,188 | 1,311 | -40% | 0 | 0 | — |
case-13 | pass→pass | 10,075 | 25,968 | +158% | 1 | 1 | 0% | 1,474 | 1,350 | -8% | 0 | 0 | — |
case-14 | pass→fail | 4,966 | 2,656 | -47% | 1 | 1 | 0% | 624 | 642 | +3% | 0 | 0 | — |
case-15 | fail→pass | 7,955 | 3,090 | -61% | 1 | 1 | 0% | 1,123 | 773 | -31% | 0 | 0 | — |
case-16 | fail→pass | 10,837 | 7,878 | -27% | 1 | 1 | 0% | 1,921 | 1,634 | -15% | 0 | 0 | — |
case-17 | fail→pass | 42,257 | 10,761 | -75% | 1 | 1 | 0% | 1,524 | 629 | -59% | 0 | 0 | — |
case-18 | fail→pass | 14,843 | 3,618 | -76% | 1 | 1 | 0% | 2,674 | 739 | -72% | 0 | 0 | — |
case-19 | fail→pass | 10,430 | 6,735 | -35% | 1 | 1 | 0% | 1,577 | 1,431 | -9% | 0 | 0 | — |
case-20 | pass→pass | 14,109 | 12,551 | -11% | 1 | 1 | 0% | 2,189 | 2,477 | +13% | 0 | 0 | — |
case-21 | fail→pass | 11,620 | 1,989 | -83% | 1 | 1 | 0% | 1,573 | 501 | -68% | 0 | 0 | — |
case-22 | fail→fail | 5,953 | 2,047 | -66% | 1 | 1 | 0% | 712 | 578 | -19% | 0 | 0 | — |
case-23 | fail→pass | 7,088 | 4,039 | -43% | 1 | 1 | 0% | 1,101 | 794 | -28% | 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 22 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 +39 percentage points is the difference between those two pass rates over the 22 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.