Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turn a curious developer into an activated one: the docs, the quickstart, and the first-run experience that gets them to their first real win fast, and back again. Use when people sign up or star the repo but never get it working, come once and never return, or you're about to pour traffic into a first-run that leaks.
.claude/skills/aidevgtm-time-to-first-value/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-02 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-05 | ✗→✓ | ▲ Improved | 28% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-12 | ✗→✓ | ▲ Improved | 42% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 43% | 0% |
> For a dev tool, the product is the pitch. A developer tries it alone and decides in minutes. Acquisition without activation is just a faster way to lose people.
Use this when: people sign up or star the repo and then nothing, they get it working once and never come back, your docs assume the reader already understands it, or you are about to spend on a launch or a channel while the first ten minutes still leak.
You can nail your positioning, your homepage, and your launch, and still lose everyone in the first ten minutes. For a dev tool the first run is the sale: the developer tries it, alone, without talking to you, and decides. Two gates, straight from Frankl's DREAM funnel:
Fix this before you scale acquisition. Pouring users into a leaky first run just loses them faster, and know-if-its-working will show you that retention, not signups, is the number that matters.
The weekend test (Frankl). A motivated developer should get your product doing something real over a weekend, without a call, a demo, or an email to you. If they cannot, that is your single highest-leverage work, above any channel, launch, or homepage tweak.
Time to first value (Czakon: "time to Hello World"). Measured in minutes, not hours, and the first meaningful win in the same sitting, not after an onboarding call next Tuesday. The first "Hello World" should be minutes in; the first real "oh, this is useful" should be that same session.
The win belongs to the developer, not your product (Frankl: the developer is the hero). The aha is "look what I just did," not "look what our platform can do." Design the first run so the developer feels capable, fast.
The core tool. Do not try to document every path. Nail the ONE most common path to first value.
Removes it:
foo and bar (Czakon: realistic sandbox data gets you most of the way).Adds it:
Can a motivated new dev reach a real win alone, in one sitting, without talking to you?
├─ NO → this is your highest-leverage work, above any channel or launch.
│ Run a friction log on the one core path and cut the top 3 barriers.
└─ YES → do they come back the next week?
├─ NO → the first win isn't valuable or sticky enough. Wrong "aha,"
│ or no reason to return. Re-pick the activation moment.
└─ YES → now acquisition is worth scaling. Go to `first-50-users`.know-if-its-working).Built from real dev-tool GTM experience, with frameworks from Adam Frankl (The Developer-Facing Startup) and Jakub Czakon (markepear.dev). When a framework can't make the call, that's what a human is for: The DevTool GTM Company.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 25,219 | 22,177 | -12% | 1 | 1 | 0% | 3,605 | 4,527 | +26% | 0 | 0 | — |
case-02 | fail→pass | 29,146 | 18,219 | -37% | 1 | 1 | 0% | 4,311 | 4,256 | -1% | 0 | 0 | — |
case-03 | fail→fail | 19,125 | 16,589 | -13% | 1 | 1 | 0% | 2,701 | 3,994 | +48% | 0 | 0 | — |
case-04 | pass→pass | 24,591 | 17,163 | -30% | 1 | 1 | 0% | 3,463 | 4,237 | +22% | 0 | 0 | — |
case-05 | fail→pass | 20,540 | 17,420 | -15% | 1 | 1 | 0% | 3,018 | 3,855 | +28% | 0 | 0 | — |
case-06 | pass→pass | 20,866 | 15,845 | -24% | 1 | 1 | 0% | 2,928 | 3,554 | +21% | 0 | 0 | — |
case-07 | pass→pass | 23,451 | 16,966 | -28% | 1 | 1 | 0% | 3,123 | 3,818 | +22% | 0 | 0 | — |
case-08 | fail→pass | 22,203 | 17,920 | -19% | 1 | 1 | 0% | 3,241 | 3,868 | +19% | 0 | 0 | — |
case-09 | fail→fail | 17,452 | 14,712 | -16% | 1 | 1 | 0% | 2,525 | 3,543 | +40% | 0 | 0 | — |
case-10 | pass→pass | 18,035 | 17,006 | -6% | 1 | 1 | 0% | 2,613 | 3,924 | +50% | 0 | 0 | — |
case-11 | fail→fail | 19,718 | 21,716 | +10% | 1 | 1 | 0% | 2,942 | 4,473 | +52% | 0 | 0 | — |
case-12 | fail→pass | 12,037 | 9,435 | -22% | 1 | 1 | 0% | 1,926 | 2,731 | +42% | 0 | 0 | — |
case-13 | pass→pass | 22,461 | 18,780 | -16% | 1 | 1 | 0% | 3,431 | 3,953 | +15% | 0 | 0 | — |
case-14 | fail→pass | 18,833 | 17,500 | -7% | 1 | 1 | 0% | 2,897 | 4,134 | +43% | 0 | 0 | — |
case-15 | pass→pass | 18,542 | 16,404 | -12% | 1 | 1 | 0% | 2,619 | 3,760 | +44% | 0 | 0 | — |
case-16 | pass→pass | 17,797 | 14,615 | -18% | 1 | 1 | 0% | 2,507 | 3,440 | +37% | 0 | 0 | — |
case-17 | pass→pass | 16,843 | 16,313 | -3% | 1 | 1 | 0% | 2,334 | 3,986 | +71% | 0 | 0 | — |
case-18 | fail→fail | 16,060 | 11,487 | -28% | 1 | 1 | 0% | 2,204 | 3,090 | +40% | 0 | 0 | — |
case-19 | fail→pass | 17,724 | 14,953 | -16% | 1 | 1 | 0% | 2,935 | 3,581 | +22% | 0 | 0 | — |
case-20 | pass→pass | 19,462 | 13,274 | -32% | 1 | 1 | 0% | 2,952 | 3,665 | +24% | 0 | 0 | — |
case-21 | pass→pass | 15,258 | 17,384 | +14% | 1 | 1 | 0% | 2,400 | 4,552 | +90% | 0 | 0 | — |
case-22 | pass→pass | 19,162 | 16,608 | -13% | 1 | 1 | 0% | 3,056 | 3,766 | +23% | 0 | 0 | — |
case-23 | fail→pass | 17,394 | 14,129 | -19% | 1 | 1 | 0% | 2,588 | 3,359 | +30% | 0 | 0 | — |
case-24 | pass→pass | 18,285 | 10,929 | -40% | 1 | 1 | 0% | 2,598 | 2,798 | +8% | 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. 24 cases were attempted. The headline lift of +29 percentage points is the difference between those two pass rates over the 24 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.