Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Build a personal library of reusable prompts for the things you ask AI again and again — so you stop rewriting the same request from scratch. Use when asked help me build a prompt library, save my best prompts, I keep writing the same prompts, or organize my AI prompts. Produces a captured set of your recurring AI tasks turned into reusable, parameterized prompt templates, an organization scheme so you can find them, guidance on what makes a prompt reusable (clear role, inputs, output format), a
.claude/skills/mohitagw15856-prompt-library-builder/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 14% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 56% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -1% | 0% |
| case-21 | ✓→✗ | ▼ Worse | 245% | 0% |
If you use AI regularly, you retype variations of the same requests constantly — the email rewriter, the meeting summarizer, the code explainer. A prompt library captures your best versions once, parameterized and organized, so you invoke them instead of reinventing them. This builds yours: identifies your recurring tasks, turns them into reusable templates, and sets up a system to store and improve them — a toolkit that gets more valuable every time you add to it.
Ask for these if not provided:
[like this], and the output format you want — so it works every time with just the specifics swapped.Your recurring tasks: the repeats, identified]. Templated (starter set) > Task name]: role/instruction] · inputs: [fill these] · output: format]. — reusable. > Task name]: …
What makes them reusable: clear role · explicit [inputs] · defined output · an example where fuzzy. Organize by: task type / domain] so you can find them. Store in: snippets / doc / saved prompts] for fast invoking. Improve: refine each as you use it; add new repeats as they emerge.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 40,039 | 21,455 | -46% | 1 | 1 | 0% | 4,967 | 3,868 | -22% | 0 | 0 | — |
case-02 | fail→fail | 22,798 | 23,985 | +5% | 1 | 1 | 0% | 2,677 | 3,738 | +40% | 0 | 0 | — |
case-03 | fail→fail | 21,216 | 16,133 | -24% | 1 | 1 | 0% | 3,495 | 3,663 | +5% | 0 | 0 | — |
case-04 | fail→fail | 21,743 | 19,255 | -11% | 1 | 1 | 0% | 2,447 | 2,877 | +18% | 0 | 0 | — |
case-05 | fail→fail | 17,450 | 19,894 | +14% | 1 | 1 | 0% | 2,897 | 3,061 | +6% | 0 | 0 | — |
case-06 | fail→pass | 19,736 | 14,116 | -28% | 1 | 1 | 0% | 2,967 | 3,377 | +14% | 0 | 0 | — |
case-07 | fail→pass | 21,188 | 20,376 | -4% | 1 | 1 | 0% | 2,238 | 3,502 | +56% | 0 | 0 | — |
case-08 | fail→fail | 22,217 | 22,022 | -1% | 1 | 1 | 0% | 2,840 | 3,924 | +38% | 0 | 0 | — |
case-09 | fail→fail | 20,918 | 20,523 | -2% | 1 | 1 | 0% | 2,493 | 3,757 | +51% | 0 | 0 | — |
case-10 | fail→pass | 21,472 | 19,888 | -7% | 1 | 1 | 0% | 2,429 | 3,402 | +40% | 0 | 0 | — |
case-11 | fail→fail | 22,212 | 20,849 | -6% | 1 | 1 | 0% | 2,913 | 3,949 | +36% | 0 | 0 | — |
case-12 | pass→pass | 24,717 | 20,126 | -19% | 1 | 1 | 0% | 2,907 | 3,421 | +18% | 0 | 0 | — |
case-13 | fail→fail | 27,840 | 20,621 | -26% | 1 | 1 | 0% | 3,307 | 3,261 | -1% | 0 | 0 | — |
case-14 | fail→fail | 26,013 | 23,470 | -10% | 1 | 1 | 0% | 2,991 | 3,705 | +24% | 0 | 0 | — |
case-15 | fail→pass | 30,026 | 23,922 | -20% | 1 | 1 | 0% | 4,024 | 3,996 | -1% | 0 | 0 | — |
case-16 | pass→pass | 22,383 | 20,413 | -9% | 1 | 1 | 0% | 2,639 | 3,724 | +41% | 0 | 0 | — |
case-17 | fail→fail | 23,006 | 24,373 | +6% | 1 | 1 | 0% | 2,695 | 3,590 | +33% | 0 | 0 | — |
case-18 | fail→fail | 26,860 | 19,860 | -26% | 1 | 1 | 0% | 3,339 | 3,585 | +7% | 0 | 0 | — |
case-19 | fail→fail | 40,394 | 23,860 | -41% | 1 | 1 | 0% | 3,001 | 3,671 | +22% | 0 | 0 | — |
case-20 | fail→fail | 15,865 | 23,189 | +46% | 1 | 1 | 0% | 1,750 | 3,297 | +88% | 0 | 0 | — |
case-21 | pass→fail | 12,973 | 24,532 | +89% | 1 | 1 | 0% | 1,020 | 3,523 | +245% | 0 | 0 | — |
case-22 | pass→pass | 14,149 | 23,410 | +65% | 1 | 1 | 0% | 1,749 | 4,288 | +145% | 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.