Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Survive (and even enjoy) small talk — how to start it, keep it going past the weather, and exit gracefully — for people who find it painful. Use when asked I'm bad at small talk, help me with small talk, what do I say at [event], or how do I make conversation. Produces conversation openers that fit the setting, the technique for keeping it flowing (curiosity, follow-up questions, the little disclosures that deepen it), how to get past surface topics toward something real, graceful exit lines, an
.claude/skills/mohitagw15856-small-talk-survival/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | 7% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 12% | 0% |
| case-13 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 33% | 0% |
| case-19 | ✓→✗ | ▼ Worse | -31% | 0% |
Small talk feels pointless and excruciating to a lot of people — but it's the bridge every real connection crosses first. The good news: it's a learnable skill, not a personality trait. This gives you openers that fit the moment, the trick to keeping conversation flowing (genuine curiosity + follow-up questions), how to steer past the weather toward something you'd actually enjoy talking about, and clean ways to exit — for the specific situation you're dreading.
Ask for these if not provided:
Openers (for this setting): context-based ways to start]. Keep it flowing: ask + follow up ("how come?" / "what was that like?") · add a little about yourself. Get past the surface: steer toward interests/excitement/what they're into]. Graceful exits: "I'm going to… — really nice talking to you"]. The reframe: it's a bridge, not the point — and they probably find it hard too.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 23,769 | 15,789 | -34% | 1 | 1 | 0% | 2,674 | 2,863 | +7% | 0 | 0 | — |
case-02 | pass→pass | 25,301 | 18,744 | -26% | 1 | 1 | 0% | 2,408 | 2,824 | +17% | 0 | 0 | — |
case-03 | pass→pass | 24,131 | 13,236 | -45% | 1 | 1 | 0% | 2,759 | 3,143 | +14% | 0 | 0 | — |
case-04 | pass→pass | 17,599 | 12,386 | -30% | 1 | 1 | 0% | 2,555 | 2,970 | +16% | 0 | 0 | — |
case-05 | fail→pass | 21,399 | 16,021 | -25% | 1 | 1 | 0% | 2,414 | 2,714 | +12% | 0 | 0 | — |
case-06 | pass→pass | 16,999 | 20,558 | +21% | 1 | 1 | 0% | 2,181 | 2,910 | +33% | 0 | 0 | — |
case-07 | pass→pass | 18,972 | 14,890 | -22% | 1 | 1 | 0% | 2,093 | 2,541 | +21% | 0 | 0 | — |
case-08 | pass→pass | 24,658 | 19,561 | -21% | 1 | 1 | 0% | 2,646 | 2,899 | +10% | 0 | 0 | — |
case-09 | pass→pass | 36,039 | 13,391 | -63% | 1 | 1 | 0% | 2,060 | 2,606 | +27% | 0 | 0 | — |
case-10 | pass→pass | 18,682 | 15,354 | -18% | 1 | 1 | 0% | 1,949 | 2,623 | +35% | 0 | 0 | — |
case-11 | pass→pass | 15,948 | 17,717 | +11% | 1 | 1 | 0% | 2,002 | 2,768 | +38% | 0 | 0 | — |
case-12 | pass→pass | 24,164 | 9,856 | -59% | 1 | 1 | 0% | 2,080 | 2,490 | +20% | 0 | 0 | — |
case-13 | fail→pass | 18,385 | 15,711 | -15% | 1 | 1 | 0% | 2,012 | 2,671 | +33% | 0 | 0 | — |
case-14 | fail→pass | 18,170 | 18,290 | +1% | 1 | 1 | 0% | 1,984 | 2,648 | +33% | 0 | 0 | — |
case-15 | pass→pass | 20,048 | 18,085 | -10% | 1 | 1 | 0% | 1,909 | 2,592 | +36% | 0 | 0 | — |
case-16 | pass→pass | 16,874 | 15,158 | -10% | 1 | 1 | 0% | 2,176 | 2,667 | +23% | 0 | 0 | — |
case-17 | pass→pass | 24,061 | 17,824 | -26% | 1 | 1 | 0% | 2,416 | 2,610 | +8% | 0 | 0 | — |
case-18 | pass→pass | 23,270 | 16,139 | -31% | 1 | 1 | 0% | 2,341 | 2,667 | +14% | 0 | 0 | — |
case-19 | pass→fail | 28,302 | 17,060 | -40% | 1 | 1 | 0% | 3,673 | 2,529 | -31% | 0 | 0 | — |
case-20 | pass→pass | 21,588 | 30,955 | +43% | 1 | 1 | 0% | 2,537 | 4,713 | +86% | 0 | 0 | — |
case-21 | pass→pass | 21,911 | 19,744 | -10% | 1 | 1 | 0% | 2,073 | 2,781 | +34% | 0 | 0 | — |
case-22 | pass→pass | 22,306 | 19,763 | -11% | 1 | 1 | 0% | 2,077 | 2,773 | +34% | 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 +14 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.