Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Open-source growth playbooks for AI products, B2B SaaS, and developer tools. Covers Product Hunt launch, GitHub star growth, KOL/UGC strategy, ASO, and community building. MIT licensed.
.claude/skills/davepoon-gingiris-growth-playbooks/SKILL.md| Model | Eval pass | Runs |
|---|---|---|
| gemini-3.6-flash | 97% | 45 |
| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -52% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -45% | 0% |
| case-03 | ✗→✓ | ▲ Improved | -66% | 0% |
| case-04 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-05 | ✗→✓ | ▲ Improved | -77% | 0% |
Battle-tested, open-source growth playbooks for developers and founders building AI products, B2B SaaS, and developer tools.
Playbook repos: github.com/Gingiris
Covers the full launch cycle: Product Hunt #1 strategy, KOL outreach, UGC campaigns, Reddit launch, and PR.
Go-to-market strategy for B2B AI products: PLG, SLG, affiliate, partner channels, and community building.
GitHub star growth strategy: README optimization, HackerNews Show HN, community seeding, and contributor flywheel.
App Store Optimization, keyword research, UGC content strategy, and cold start for mobile apps.
Reference the relevant playbook for your project type. All playbooks are MIT licensed and free to use, adapt, and contribute to.
None — these are markdown-based strategy documents, no MCP or API keys required.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 15,206 | 3,833 | -75% | 1 | 1 | 0% | 2,217 | 1,068 | -52% | 0 | 0 | — |
case-02 | fail→pass | 16,831 | 7,400 | -56% | 1 | 1 | 0% | 2,917 | 1,594 | -45% | 0 | 0 | — |
case-03 | fail→pass | 19,199 | 3,805 | -80% | 1 | 1 | 0% | 2,934 | 987 | -66% | 0 | 0 | — |
case-04 | fail→pass | 21,045 | 4,076 | -81% | 1 | 1 | 0% | 3,076 | 1,100 | -64% | 0 | 0 | — |
case-05 | fail→pass | 20,509 | 2,203 | -89% | 1 | 1 | 0% | 3,433 | 800 | -77% | 0 | 0 | — |
case-06 | fail→pass | 19,577 | 4,167 | -79% | 1 | 1 | 0% | 3,006 | 1,160 | -61% | 0 | 0 | — |
case-07 | fail→pass | 14,255 | 3,270 | -77% | 1 | 1 | 0% | 2,237 | 891 | -60% | 0 | 0 | — |
case-08 | fail→pass | 19,435 | 3,571 | -82% | 1 | 1 | 0% | 2,620 | 1,021 | -61% | 0 | 0 | — |
case-09 | fail→pass | 14,321 | 2,481 | -83% | 1 | 1 | 0% | 2,059 | 819 | -60% | 0 | 0 | — |
case-10 | fail→pass | 11,111 | 2,059 | -81% | 1 | 1 | 0% | 1,865 | 702 | -62% | 0 | 0 | — |
case-11 | fail→pass | 5,986 | 1,762 | -71% | 1 | 1 | 0% | 812 | 607 | -25% | 0 | 0 | — |
case-12 | pass→pass | 4,408 | 2,339 | -47% | 1 | 1 | 0% | 623 | 703 | +13% | 0 | 0 | — |
case-17 | fail→pass | 7,535 | 2,039 | -73% | 1 | 1 | 0% | 1,188 | 645 | -46% | 0 | 0 | — |
case-13 | fail→pass | 17,664 | 4,813 | -73% | 1 | 1 | 0% | 2,544 | 1,139 | -55% | 0 | 0 | — |
case-14 | fail→pass | 15,036 | 3,804 | -75% | 1 | 1 | 0% | 2,164 | 1,084 | -50% | 0 | 0 | — |
case-15 | fail→pass | 15,500 | 2,633 | -83% | 1 | 1 | 0% | 2,409 | 833 | -65% | 0 | 0 | — |
case-16 | fail→pass | 14,055 | 2,366 | -83% | 1 | 1 | 0% | 2,528 | 716 | -72% | 0 | 0 | — |
case-18 | fail→pass | 15,842 | 4,061 | -74% | 1 | 1 | 0% | 2,570 | 1,150 | -55% | 0 | 0 | — |
case-19 | fail→pass | 8,103 | 2,929 | -64% | 1 | 1 | 0% | 1,254 | 940 | -25% | 0 | 0 | — |
case-20 | pass→pass | 11,098 | 12,735 | +15% | 1 | 1 | 0% | 1,963 | 2,736 | +39% | 0 | 0 | — |
case-21 | pass→pass | 6,013 | 3,789 | -37% | 1 | 1 | 0% | 1,025 | 1,160 | +13% | 0 | 0 | — |
case-22 | pass→pass | 4,278 | 3,926 | -8% | 1 | 1 | 0% | 593 | 1,057 | +78% | 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 +82 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.