Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Drive an email inbox to zero through a computer-use or tool-using agent — triage every message into act/delegate/defer/archive with drafts prepared, never a send. Use when asked to get my inbox to zero, triage my email hands-on, process my inbox for me, or run inbox zero. Produces the triage ledger, prepared reply drafts, and an approval-gated action plan the agent then executes read-mostly.
.claude/skills/mohitagw15856-inbox-zero-operator/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 810% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 243% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 74% | 0% |
| case-16 | ✗→✓ | ▲ Improved | 60% | 0% |
| case-17 | ✗→✓ | ▲ Improved | 235% | 0% |
The judgment layer for an agent that can actually touch a mailbox. The thinking half (what deserves you, what doesn't) and the acting half (archive, label, draft) are strictly separated: analysis is free, action is gated, and sending is never on the menu. (For triage-as-advice without tool access, email-triage covers you.)
Ask for these if not provided:
| Thread | From | Class | Why (one line) | Action | |---|---|---|---|---| Drafts prepared: n] · To archive: n] · Deferred: n with dates] Needs you (the 🔴 list, ranked): …
For computer-use or tool-using agents with mailbox access (Gmail/Outlook APIs or UI). Runtimes without tools deliver the ledger as a document. Rules per SKILLSPEC.md §5.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 11,626 | 18,312 | +58% | 1 | 1 | 0% | 2,053 | 4,187 | +104% | 0 | 0 | — |
case-02 | fail→pass | 4,788 | 13,075 | +173% | 1 | 1 | 0% | 386 | 3,512 | +810% | 0 | 0 | — |
case-03 | pass→pass | 15,173 | 13,042 | -14% | 1 | 1 | 0% | 2,706 | 3,610 | +33% | 0 | 0 | — |
case-04 | fail→pass | 3,736 | 6,510 | +74% | 1 | 1 | 0% | 592 | 2,033 | +243% | 0 | 0 | — |
case-05 | pass→pass | 8,975 | 7,994 | -11% | 1 | 1 | 0% | 1,779 | 2,368 | +33% | 0 | 0 | — |
case-06 | pass→pass | 6,914 | 7,019 | +2% | 1 | 1 | 0% | 1,342 | 2,307 | +72% | 0 | 0 | — |
case-07 | pass→pass | 6,940 | 5,977 | -14% | 1 | 1 | 0% | 1,084 | 1,932 | +78% | 0 | 0 | — |
case-08 | pass→pass | 3,946 | 3,831 | -3% | 1 | 1 | 0% | 649 | 1,670 | +157% | 0 | 0 | — |
case-09 | pass→pass | 5,516 | 3,354 | -39% | 1 | 1 | 0% | 878 | 1,609 | +83% | 0 | 0 | — |
case-10 | pass→pass | 10,005 | 3,510 | -65% | 1 | 1 | 0% | 1,600 | 1,655 | +3% | 0 | 0 | — |
case-11 | pass→pass | 9,742 | 4,257 | -56% | 1 | 1 | 0% | 1,528 | 1,684 | +10% | 0 | 0 | — |
case-12 | pass→pass | 7,187 | 4,719 | -34% | 1 | 1 | 0% | 1,208 | 1,712 | +42% | 0 | 0 | — |
case-13 | fail→pass | 5,427 | 2,550 | -53% | 1 | 1 | 0% | 791 | 1,375 | +74% | 0 | 0 | — |
case-14 | pass→pass | 4,708 | 4,421 | -6% | 1 | 1 | 0% | 792 | 1,740 | +120% | 0 | 0 | — |
case-15 | fail→fail | 6,166 | 1,933 | -69% | 1 | 1 | 0% | 941 | 1,271 | +35% | 0 | 0 | — |
case-16 | fail→pass | 6,138 | 3,590 | -42% | 1 | 1 | 0% | 973 | 1,561 | +60% | 0 | 0 | — |
case-17 | fail→pass | 3,372 | 4,159 | +23% | 1 | 1 | 0% | 518 | 1,733 | +235% | 0 | 0 | — |
case-18 | fail→fail | 7,179 | 3,982 | -45% | 1 | 1 | 0% | 1,309 | 1,695 | +29% | 0 | 0 | — |
case-19 | pass→pass | 12,048 | 4,717 | -61% | 1 | 1 | 0% | 1,805 | 1,803 | -0% | 0 | 0 | — |
case-20 | pass→pass | 4,021 | 3,328 | -17% | 1 | 1 | 0% | 715 | 1,549 | +117% | 0 | 0 | — |
case-21 | pass→pass | 3,617 | 3,046 | -16% | 1 | 1 | 0% | 568 | 1,488 | +162% | 0 | 0 | — |
case-22 | fail→fail | 9,417 | 4,214 | -55% | 1 | 1 | 0% | 1,590 | 1,700 | +7% | 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 +23 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.