Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Use when an agent needs to write user stories for a project
.claude/skills/nicepkg-writing-user-stories/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | -64% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -74% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -54% | 0% |
| case-03 | ✓→✗ | ▼ Worse | 12% | 0% |
| case-16 | ✓→✗ | ▼ Worse | -52% | 0% |
Help Claude and subagents write properly formatted user stories for task definition.
Core principle: User stories are formulaic ways of expressing user requirements. They identify the perosna of the actor, what they want to do, and what benefit are they hoping to gain from their actions.
Use this skill when:
As a PERSONA I want to DESCRIPTION_OF_ACTION So that DESCRIPTION_OF_BENEFIT.
As a non-technical stakeholder, I want to understand what changed in a release without reading code diffs so that I can communicate updates to customers effectively.
As a new user, I want to see examples of AI-generated code before connecting my repository so that I understand the value proposition before committing to integration.
As a home cook, I want to see that recipes have been tested by professional chefs so that I can know whether they are worth trying.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-03 | pass→fail | 14,742 | 14,184 | -4% | 1 | 1 | 0% | 2,206 | 2,477 | +12% | 0 | 0 | — |
case-01 | pass→pass | 14,682 | 12,738 | -13% | 1 | 1 | 0% | 2,891 | 1,865 | -35% | 0 | 0 | — |
case-02 | pass→pass | 8,952 | 7,875 | -12% | 1 | 1 | 0% | 1,393 | 1,338 | -4% | 0 | 0 | — |
case-04 | pass→pass | 8,806 | 2,079 | -76% | 1 | 1 | 0% | 1,411 | 591 | -58% | 0 | 0 | — |
case-05 | pass→pass | 7,866 | 3,560 | -55% | 1 | 1 | 0% | 1,299 | 782 | -40% | 0 | 0 | — |
case-06 | fail→pass | 12,044 | 2,889 | -76% | 1 | 1 | 0% | 1,916 | 682 | -64% | 0 | 0 | — |
case-07 | pass→pass | 11,536 | 2,184 | -81% | 1 | 1 | 0% | 1,816 | 573 | -68% | 0 | 0 | — |
case-08 | pass→pass | 7,347 | 3,280 | -55% | 1 | 1 | 0% | 1,172 | 715 | -39% | 0 | 0 | — |
case-09 | pass→pass | 8,613 | 2,450 | -72% | 1 | 1 | 0% | 1,361 | 590 | -57% | 0 | 0 | — |
case-10 | fail→pass | 15,767 | 2,112 | -87% | 1 | 1 | 0% | 2,146 | 560 | -74% | 0 | 0 | — |
case-11 | pass→pass | 9,334 | 2,773 | -70% | 1 | 1 | 0% | 1,407 | 652 | -54% | 0 | 0 | — |
case-12 | pass→pass | 13,465 | 3,691 | -73% | 1 | 1 | 0% | 1,902 | 771 | -59% | 0 | 0 | — |
case-13 | pass→pass | 8,329 | 2,539 | -70% | 1 | 1 | 0% | 1,290 | 678 | -47% | 0 | 0 | — |
case-14 | pass→pass | 7,624 | 2,430 | -68% | 1 | 1 | 0% | 1,337 | 631 | -53% | 0 | 0 | — |
case-15 | pass→pass | 10,770 | 2,895 | -73% | 1 | 1 | 0% | 1,722 | 668 | -61% | 0 | 0 | — |
case-16 | pass→fail | 7,998 | 2,531 | -68% | 1 | 1 | 0% | 1,309 | 631 | -52% | 0 | 0 | — |
case-17 | fail→pass | 10,060 | 3,152 | -69% | 1 | 1 | 0% | 1,568 | 725 | -54% | 0 | 0 | — |
case-18 | pass→pass | 8,032 | 2,219 | -72% | 1 | 1 | 0% | 1,321 | 591 | -55% | 0 | 0 | — |
case-19 | pass→fail | 9,597 | 2,122 | -78% | 1 | 1 | 0% | 1,559 | 549 | -65% | 0 | 0 | — |
case-20 | pass→pass | 8,833 | 2,218 | -75% | 1 | 1 | 0% | 1,477 | 559 | -62% | 0 | 0 | — |
case-21 | pass→pass | 8,506 | 2,456 | -71% | 1 | 1 | 0% | 1,366 | 580 | -58% | 0 | 0 | — |
case-22 | pass→pass | 9,525 | 3,112 | -67% | 1 | 1 | 0% | 1,500 | 645 | -57% | 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 0 percentage points is the difference between those two pass rates over the 22 comparable cases. 3 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.