Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prepare the user for a REAL upcoming meeting by pulling the actual Calendar event, its attendees, the linked Drive docs, and the last email/Slack thread — then producing a brief. Use when asked to prep me for my next meeting, get me ready for the 2pm, or what do I need for the sync with X in Cowork. Reads the event via the Google Calendar connector, gathers the attached and related material via Drive/Gmail, and produces a one-page meeting-brief artifact with objective, context, open threads, and
.claude/skills/mohitagw15856-meeting-prep-live/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 30% | 0% |
| case-04 | ✗→✓ | ▲ Improved | 124% | 0% |
| case-06 | ✗→✓ | ▲ Improved | 63% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 34% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 56% | 0% |
Walking into a meeting cold costs the first ten minutes. In Claude Cowork this skill assembles the brief from the user's real calendar and files — the event, who's in the room, what's attached, and where the last conversation left off — so they arrive with the context already in hand.
Ask for these if not provided:
Guardrails: cite the real source for every claim; never fabricate an attendee position or a decision that isn't in the material; if a connector is unauthorised, produce the brief from what's available and name the gap.
A Meeting Brief artifact:
Objective: one line · Decision needed: the one outcome
| Person | Stake / likely position | |---|---|
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 12,836 | 11,066 | -14% | 1 | 1 | 0% | 2,058 | 2,666 | +30% | 0 | 0 | — |
case-02 | fail→fail | 5,296 | 5,582 | +5% | 1 | 1 | 0% | 879 | 1,826 | +108% | 0 | 0 | — |
case-03 | fail→fail | 17,535 | 10,732 | -39% | 1 | 1 | 0% | 2,837 | 2,548 | -10% | 0 | 0 | — |
case-04 | fail→pass | 7,013 | 9,392 | +34% | 1 | 1 | 0% | 1,048 | 2,347 | +124% | 0 | 0 | — |
case-05 | pass→pass | 5,074 | 6,167 | +22% | 1 | 1 | 0% | 819 | 1,812 | +121% | 0 | 0 | — |
case-06 | fail→pass | 7,810 | 8,048 | +3% | 1 | 1 | 0% | 1,315 | 2,137 | +63% | 0 | 0 | — |
case-07 | fail→pass | 11,903 | 11,148 | -6% | 1 | 1 | 0% | 1,884 | 2,530 | +34% | 0 | 0 | — |
case-08 | pass→pass | 17,085 | 9,802 | -43% | 1 | 1 | 0% | 2,639 | 2,274 | -14% | 0 | 0 | — |
case-09 | pass→pass | 6,212 | 8,589 | +38% | 1 | 1 | 0% | 1,004 | 2,208 | +120% | 0 | 0 | — |
case-10 | pass→pass | 12,094 | 9,872 | -18% | 1 | 1 | 0% | 2,053 | 2,416 | +18% | 0 | 0 | — |
case-11 | pass→pass | 13,347 | 9,716 | -27% | 1 | 1 | 0% | 2,227 | 2,352 | +6% | 0 | 0 | — |
case-12 | pass→pass | 20,655 | 9,179 | -56% | 1 | 1 | 0% | 1,739 | 2,309 | +33% | 0 | 0 | — |
case-13 | fail→fail | 16,287 | 18,607 | +14% | 1 | 1 | 0% | 779 | 1,566 | +101% | 0 | 0 | — |
case-14 | fail→pass | 12,164 | 11,650 | -4% | 1 | 1 | 0% | 1,061 | 1,655 | +56% | 0 | 0 | — |
case-15 | fail→pass | 8,840 | 21,235 | +140% | 1 | 1 | 0% | 1,425 | 1,902 | +33% | 0 | 0 | — |
case-16 | fail→pass | 18,618 | 19,249 | +3% | 1 | 1 | 0% | 2,091 | 2,495 | +19% | 0 | 0 | — |
case-17 | pass→pass | 13,056 | 19,962 | +53% | 1 | 1 | 0% | 2,151 | 2,580 | +20% | 0 | 0 | — |
case-18 | fail→pass | 11,553 | 9,211 | -20% | 1 | 1 | 0% | 2,026 | 2,493 | +23% | 0 | 0 | — |
case-19 | pass→pass | 8,476 | 5,927 | -30% | 1 | 1 | 0% | 1,443 | 1,930 | +34% | 0 | 0 | — |
case-20 | pass→pass | 4,924 | 8,041 | +63% | 1 | 1 | 0% | 905 | 2,068 | +129% | 0 | 0 | — |
case-21 | pass→pass | 4,916 | 4,277 | -13% | 1 | 1 | 0% | 762 | 1,540 | +102% | 0 | 0 | — |
case-22 | pass→pass | 3,714 | 3,059 | -18% | 1 | 1 | 0% | 600 | 1,407 | +135% | 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 +36 percentage points is the difference between those two pass rates over the 22 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.