Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Research a topic from the last 30 days on Reddit + X + Web, become an expert, and write copy-paste-ready prompts for the user's target tool.
.claude/skills/dokhacgiakhoa-last30days/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | 27% | 0% |
| case-11 | ✓→✗ | ▼ Worse | -48% | 0% |
| case-07 | ✓→✓ | ＝ Same ✓ | 72% | 0% |
| case-19 | ✓→✓ | ＝ Same ✓ | 59% | 0% |
| case-20 | ✓→✓ | ＝ Same ✓ | 200% | 0% |
Research ANY topic across Reddit, X, and the web. Surface what people are actually discussing, recommending, and debating right now.
Use cases:
Before doing anything, parse the user's input for:
Common patterns:
[topic] for [tool] → "web mockups for Nano Banana Pro" → TOOL IS SPECIFIED[topic] prompts for [tool] → "UI design prompts for Midjourney" → TOOL IS SPECIFIED[topic] → "iOS design mockups" → TOOL NOT SPECIFIED, that's OKIMPORTANT: Do NOT ask about target tool before research.
Store these variables:
TOPIC = [extracted topic]TARGET_TOOL = [extracted tool, or "unknown" if not specified]QUERY_TYPE = [RECOMMENDATIONS | NEWS | HOW-TO | GENERAL]The skill works in three modes based on available API keys:
API keys are OPTIONAL. The skill will work without them using WebSearch fallback.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 19,308 | 20,357 | +5% | 1 | 1 | 0% | 2,377 | 1,303 | -45% | 0 | 0 | — |
case-02 | fail→pass | 31,260 | 31,160 | -0% | 1 | 1 | 0% | 2,915 | 3,698 | +27% | 0 | 0 | — |
case-03 | fail→fail | 13,793 | 10,380 | -25% | 1 | 1 | 0% | 2,249 | 1,609 | -28% | 0 | 0 | — |
case-04 | fail→fail | 22,367 | 18,157 | -19% | 1 | 1 | 0% | 2,785 | 1,303 | -53% | 0 | 0 | — |
case-05 | fail→fail | 12,593 | 15,247 | +21% | 1 | 1 | 0% | 2,104 | 1,301 | -38% | 0 | 0 | — |
case-06 | fail→fail | 21,401 | 20,135 | -6% | 1 | 1 | 0% | 2,721 | 1,495 | -45% | 0 | 0 | — |
case-07 | pass→pass | 18,266 | 27,580 | +51% | 1 | 1 | 0% | 2,246 | 3,862 | +72% | 0 | 0 | — |
case-08 | fail→fail | 21,959 | 46,488 | +112% | 1 | 1 | 0% | 2,924 | 1,372 | -53% | 0 | 0 | — |
case-09 | fail→fail | 11,510 | 17,983 | +56% | 1 | 1 | 0% | 395 | 1,329 | +236% | 0 | 0 | — |
case-10 | fail→fail | 17,847 | 13,810 | -23% | 1 | 1 | 0% | 361 | 1,464 | +306% | 0 | 0 | — |
case-11 | pass→fail | 23,210 | 13,343 | -43% | 1 | 1 | 0% | 2,468 | 1,291 | -48% | 0 | 0 | — |
case-12 | fail→fail | 18,477 | 12,838 | -31% | 1 | 1 | 0% | 2,141 | 1,450 | -32% | 0 | 0 | — |
case-13 | fail→fail | 19,516 | 12,864 | -34% | 1 | 1 | 0% | 2,393 | 1,419 | -41% | 0 | 0 | — |
case-14 | fail→fail | 19,099 | 17,359 | -9% | 1 | 1 | 0% | 2,410 | 1,352 | -44% | 0 | 0 | — |
case-15 | fail→fail | 21,174 | 11,829 | -44% | 1 | 1 | 0% | 2,379 | 1,334 | -44% | 0 | 0 | — |
case-16 | fail→fail | 15,288 | 12,093 | -21% | 1 | 1 | 0% | 2,371 | 1,194 | -50% | 0 | 0 | — |
case-17 | fail→fail | 18,250 | 19,872 | +9% | 1 | 1 | 0% | 2,137 | 3,561 | +67% | 0 | 0 | — |
case-18 | fail→fail | 25,452 | 27,476 | +8% | 1 | 1 | 0% | 3,678 | 1,424 | -61% | 0 | 0 | — |
case-19 | pass→pass | 20,626 | 24,536 | +19% | 1 | 1 | 0% | 2,823 | 4,500 | +59% | 0 | 0 | — |
case-20 | pass→pass | 2,834 | 8,628 | +204% | 1 | 1 | 0% | 503 | 1,508 | +200% | 0 | 0 | — |
case-21 | pass→pass | 25,901 | 24,580 | -5% | 1 | 1 | 0% | 4,885 | 6,501 | +33% | 0 | 0 | — |
case-22 | fail→fail | 23,217 | 14,244 | -39% | 1 | 1 | 0% | 2,400 | 1,336 | -44% | 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 6 counted toward the lift figure. The other 16 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 0 percentage points is the difference between those two pass rates over the 6 comparable cases. 12 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.