Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Prepare for a specific talk, presentation, or speech — a clear structure, a strong open and close, delivery and nerves handling, and a rehearsal plan — so you land it. Use when asked to help me prepare a talk/presentation/speech, prep for public speaking, I have to give a presentation, or calm my speaking nerves. Produces a message-first structure built on your core point and audience, a memorable opening and closing, delivery guidance (pace, pauses, notes vs script), a nerves-management plan, a
.claude/skills/mohitagw15856-public-speaking-prep/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -12% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -16% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 36% | 0% |
| case-05 | ✓→✗ | ▼ Worse | -18% | 0% |
| case-02 | ✓→✓ | = Same ✓ | 3% | 0% |
A good talk isn't about being a naturally gifted speaker — it's preparation: a clear core message, a structure the audience can follow, a strong open and close, and enough rehearsal that nerves have less room. This preps your specific talk end to end, from what you're actually trying to say to handling the shaky-hands moment before you start.
Ask for these if not provided:
Core message: the one thing they should remember]. Structure: open → point 1 · 2 · 3 with support] → close. Open: attention-earning hook]. Close: memorable takeaway]. Delivery: pace/pauses · notes vs script · slides supportive]. Nerves: prep + in-the-moment techniques]. Rehearse: out loud, timed, to a person/recorded]. Q&A: likely questions + handling the unknown ones].
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 54,750 | 30,555 | -44% | 1 | 1 | 0% | 5,606 | 4,928 | -12% | 0 | 0 | — |
case-02 | pass→pass | 31,908 | 60,494 | +90% | 1 | 1 | 0% | 3,986 | 4,098 | +3% | 0 | 0 | — |
case-03 | fail→pass | 74,864 | 26,168 | -65% | 1 | 1 | 0% | 5,009 | 4,210 | -16% | 0 | 0 | — |
case-04 | pass→pass | 49,471 | 35,554 | -28% | 1 | 1 | 0% | 3,789 | 5,369 | +42% | 0 | 0 | — |
case-05 | pass→fail | 57,509 | 31,236 | -46% | 1 | 1 | 0% | 8,215 | 6,729 | -18% | 0 | 0 | — |
case-06 | pass→pass | 22,388 | 29,452 | +32% | 1 | 1 | 0% | 2,462 | 3,271 | +33% | 0 | 0 | — |
case-07 | pass→pass | 19,173 | 19,957 | +4% | 1 | 1 | 0% | 1,933 | 3,127 | +62% | 0 | 0 | — |
case-08 | pass→pass | 24,072 | 48,389 | +101% | 1 | 1 | 0% | 1,804 | 3,295 | +83% | 0 | 0 | — |
case-09 | fail→fail | 21,942 | 17,430 | -21% | 1 | 1 | 0% | 2,507 | 3,222 | +29% | 0 | 0 | — |
case-10 | pass→pass | 17,242 | 10,976 | -36% | 1 | 1 | 0% | 2,072 | 2,585 | +25% | 0 | 0 | — |
case-11 | pass→pass | 16,766 | 16,049 | -4% | 1 | 1 | 0% | 1,796 | 3,065 | +71% | 0 | 0 | — |
case-12 | fail→pass | 18,267 | 18,642 | +2% | 1 | 1 | 0% | 2,579 | 3,496 | +36% | 0 | 0 | — |
case-13 | pass→pass | 28,012 | 11,966 | -57% | 1 | 1 | 0% | 2,355 | 2,635 | +12% | 0 | 0 | — |
case-14 | pass→pass | 16,548 | 15,286 | -8% | 1 | 1 | 0% | 2,050 | 3,009 | +47% | 0 | 0 | — |
case-15 | pass→pass | 31,103 | 24,935 | -20% | 1 | 1 | 0% | 2,594 | 3,823 | +47% | 0 | 0 | — |
case-16 | pass→pass | 14,819 | 15,023 | +1% | 1 | 1 | 0% | 1,873 | 2,842 | +52% | 0 | 0 | — |
case-17 | pass→pass | 16,859 | 6,686 | -60% | 1 | 1 | 0% | 2,091 | 2,025 | -3% | 0 | 0 | — |
case-18 | pass→pass | 19,069 | 19,637 | +3% | 1 | 1 | 0% | 2,874 | 3,550 | +24% | 0 | 0 | — |
case-19 | pass→pass | 13,807 | 13,690 | -1% | 1 | 1 | 0% | 2,092 | 2,990 | +43% | 0 | 0 | — |
case-20 | pass→pass | 13,800 | 12,714 | -8% | 1 | 1 | 0% | 2,133 | 2,824 | +32% | 0 | 0 | — |
case-21 | pass→pass | 21,612 | 18,627 | -14% | 1 | 1 | 0% | 2,807 | 3,706 | +32% | 0 | 0 | — |
case-22 | pass→pass | 14,075 | 14,971 | +6% | 1 | 1 | 0% | 1,870 | 3,123 | +67% | 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 +9 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.