Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Delegate so the work comes back right the first time — the brief that transfers outcome, context, and constraints (not just the task), the autonomy level stated explicitly, and the check-in design that catches drift without hovering. Use when asked hand this off properly, my delegations come back wrong, write a brief for this task I'm giving away, or how much detail do I give. Produces the delegation brief with the outcome and guardrails, the autonomy level, the check-in points, and the question
.claude/skills/mohitagw15856-delegation-brief/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 37% | 0% |
| case-02 | ✗→✓ | ▲ Improved | 1% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 57% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 32% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 59% | 0% |
Delegation fails at the handoff, not the execution: the task transferred without its outcome ("send me a draft" — of what, for whom, to achieve what?), without its constraints (discovered as rejections later), without the autonomy level (so the delegate either hovers back with every micro-decision or runs confidently off the map). The brief fixes the transfer: the outcome with its done-test, the context that makes judgment possible (why this matters, who it's for), the guardrails (the true constraints, and only the true ones), the stated autonomy level, and check-ins placed where drift is cheapest to catch — at the first artifact, not the final one.
Ask for these if not provided:
Outcome: what done looks like + the done-test] Why/for whom: the two context sentences] Guardrails (真 constraints): the true limits] · Preferences (optional, labeled): …] Autonomy: per decision class: decide-and-do / decide-tell / propose-first] Deadline: date + hard/soft stated] · Resources: access, prior art, the person to ask about X]
First-artifact review at (date) · milestone (if long) · the questions-welcome contract, said]
The delegate states outcome + first step — gaps patched now, not at delivery]
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 18,538 | 15,090 | -19% | 1 | 1 | 0% | 2,799 | 3,846 | +37% | 0 | 0 | — |
case-02 | fail→pass | 20,934 | 14,034 | -33% | 1 | 1 | 0% | 3,383 | 3,422 | +1% | 0 | 0 | — |
case-03 | pass→pass | 13,976 | 14,036 | +0% | 1 | 1 | 0% | 2,108 | 3,454 | +64% | 0 | 0 | — |
case-04 | pass→fail | 16,340 | 14,385 | -12% | 1 | 1 | 0% | 2,244 | 3,487 | +55% | 0 | 0 | — |
case-05 | pass→fail | 12,763 | 25,190 | +97% | 1 | 1 | 0% | 2,068 | 4,986 | +141% | 0 | 0 | — |
case-06 | pass→pass | 13,482 | 15,514 | +15% | 1 | 1 | 0% | 2,252 | 3,790 | +68% | 0 | 0 | — |
case-07 | pass→pass | 11,574 | 18,198 | +57% | 1 | 1 | 0% | 1,782 | 3,863 | +117% | 0 | 0 | — |
case-13 | fail→pass | 12,605 | 12,580 | -0% | 1 | 1 | 0% | 2,157 | 3,380 | +57% | 0 | 0 | — |
case-08 | pass→pass | 9,467 | 15,537 | +64% | 1 | 1 | 0% | 1,529 | 3,712 | +143% | 0 | 0 | — |
case-09 | pass→pass | 11,394 | 16,114 | +41% | 1 | 1 | 0% | 1,900 | 3,792 | +100% | 0 | 0 | — |
case-10 | pass→pass | 15,568 | 15,399 | -1% | 1 | 1 | 0% | 2,488 | 3,463 | +39% | 0 | 0 | — |
case-11 | pass→pass | 16,573 | 15,841 | -4% | 1 | 1 | 0% | 2,465 | 3,773 | +53% | 0 | 0 | — |
case-12 | pass→pass | 12,833 | 13,467 | +5% | 1 | 1 | 0% | 1,882 | 3,366 | +79% | 0 | 0 | — |
case-14 | fail→pass | 14,572 | 10,813 | -26% | 1 | 1 | 0% | 2,466 | 3,256 | +32% | 0 | 0 | — |
case-15 | pass→pass | 15,242 | 12,548 | -18% | 1 | 1 | 0% | 2,370 | 3,326 | +40% | 0 | 0 | — |
case-16 | fail→pass | 12,214 | 11,255 | -8% | 1 | 1 | 0% | 1,986 | 3,167 | +59% | 0 | 0 | — |
case-17 | pass→pass | 15,129 | 17,765 | +17% | 1 | 1 | 0% | 2,397 | 4,064 | +70% | 0 | 0 | — |
case-18 | pass→pass | 14,558 | 16,447 | +13% | 1 | 1 | 0% | 2,262 | 3,726 | +65% | 0 | 0 | — |
case-19 | fail→pass | 14,448 | 12,758 | -12% | 1 | 1 | 0% | 2,267 | 3,408 | +50% | 0 | 0 | — |
case-20 | fail→pass | 8,616 | 12,816 | +49% | 1 | 1 | 0% | 1,375 | 3,259 | +137% | 0 | 0 | — |
case-21 | pass→pass | 13,571 | 12,793 | -6% | 1 | 1 | 0% | 2,225 | 3,281 | +47% | 0 | 0 | — |
case-22 | pass→pass | 21,710 | 12,701 | -41% | 1 | 1 | 0% | 2,136 | 3,206 | +50% | 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 +23 percentage points is the difference between those two pass rates over the 22 comparable cases. 2 cases got worse with the skill loaded, and they are 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.