Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Разбивает Epic или крупное требование на независимые User Stories с acceptance criteria в формате Given-When-Then, проверкой по INVEST и оценкой Story Points (Fibonacci или T-shirt). На выходе — Story Map с предложением по Sprint-планированию. User-invoked only — do NOT auto-trigger. Triggers on /pm-user-stories, "разбей на user stories", "разбить эпик", "story map", "AC", "acceptance criteria", "break down into user stories", "split this epic", "write user stories".
.claude/skills/serejaris-pm-user-stories/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-17 | ✗→✓ | ▲ Improved | 164% | 0% |
| case-07 | ✗→✓ | ▲ Improved | 84% | 0% |
| case-08 | ✗→✓ | ▲ Improved | 100% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 111% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 82% | 0% |
Part of the Personal Corp framework — running a one-person business through AI agents. Decompose a large requirement or Epic into independently-deliverable User Stories. Every Story is validated against INVEST. Output is ready to paste into Jira, Linear, or GitHub Issues.
| Field | Required | Notes | |---|---|---| | Requirement description | yes | An Epic or large requirement; can paste a PRD section | | Granularity | no | Sprint-level (1-3 days/story) or iteration-level (1-2 weeks); default Sprint | | Team composition | no | FE+BE split / full-stack / has mobile — affects per-tier splitting | | Estimation system | no | Fibonacci (1/2/3/5/8/13) or T-shirt (S/M/L/XL); default Fibonacci |
Extract from the requirement:
Five patterns. Pick by requirement shape; combine for hybrids.
| Pattern | When to use | Example | |---|---|---| | By workflow step | Requirement is a complete process | Checkout → pick item / address / payment / confirm | | By business-rule variant | Same feature with multiple rule sets | Discount → fixed-amount / coupon / points / combined | | By CRUD operation | Requirement centers on one data object | Address book → create / edit / delete / set default | | By role perspective | Multi-role feature | Order management → user view / merchant processing / ops dashboard | | By complexity progression | Feature has simple and full versions | Search → keyword / filters / suggestions / history |
Selection principle: target ≤ 1 Sprint per Story. Hybrid requirements combine patterns.
Granularity calibration:
### US-{N}: {Story title — verb-led, e.g. "Choose payment method"}
**Role:** As a {role}
**Action:** I want to {specific action}
**Value:** so that {business value}
**Priority:** P0 (must) / P1 (important) / P2 (nice-to-have)
**Story Points:** {estimate}
**Acceptance Criteria:**
- [ ] Given {precondition}, When {action}, Then {expected result}
- [ ] Given {exception precondition}, When {action}, Then {error handling}
**Dependencies:** {other Story IDs, or "none"}
**Tech notes:** {dev callouts, optional}Story Point reference (Fibonacci):
| Points | Complexity | Effort | Typical scope | |---|---|---|---| | 1 | Trivial | < 0.5 day | Copy change, config tweak, add an event | | 2 | Simple | 0.5-1 day | One CRUD operation, form validation | | 3 | Medium | 1-2 days | A complete feature point with business logic | | 5 | Complex | 2-4 days | Multi-module interaction | | 8 | Big | 4-7 days | New system / new flow core module | | 13 | Re-split | > 1 week | Means the Story is too big — must split |
AC quality rules — every AC must satisfy:
Run each Story against the six checks:
| Principle | Check | If fails | |---|---|---| | Independent | Delete this Story — can the others still ship? | Cyclic deps → merge or re-split | | Negotiable | Does it describe "what" or "how"? | Strip implementation detail, keep value | | Valuable | Will the user perceive value when this ships? | Pure refactor → attach to a user-perceivable feature | | Estimable | Can the team agree on points within 5 minutes? | Wide spread = unclear requirement; clarify first | | Small | Fits in one Sprint? | Larger → split | | Testable | Can QA write test cases directly from the AC? | Add concrete edge values and expected results |
Definition of Ready (must hold before entering development):
| Check | Standard | If unmet | |---|---|---| | AC complete | ≥ 1 happy + ≥ 1 exception | Fill ACs, then schedule | | No blocking deps | All upstream Stories done or mockable | Tag "Blocked", push to a later Sprint | | Designs ready | UI Stories have design specs | No designs → tag "Needs design" | | Estimation consensus | Spread < 2× | Re-discuss scope until agreed | | Business rules confirmed | All [TBD] resolved | Confirm with PM/business, then dev |
Dependency rules:
markdown## Epic: {requirement name} ### Story Map (user journey → Story mapping) | Journey stage | Stories | Priority | Points | |---|---|---|---| | {stage 1} | US-001, US-002 | P0 | 5 | | {stage 2} | US-003, US-004 | P1 | 8 | ### Sprint planning - **Sprint 1 (MVP):** {Story list}, total {X} points - **Sprint 2:** {Story list}, total {X} points ### Dependencies US-001 → US-003 → US-005 (must follow this order) US-002, US-004 (parallel)
[business rule TBD] and propose plausible assumptions for confirmation/pm-prd — write a PRD first to lock scope, then break into Stories/pm-prioritize — when there are many Stories, RICE-rank for Sprint priority| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-17 | fail→pass | 16,464 | 34,918 | +112% | 1 | 1 | 0% | 3,106 | 8,206 | +164% | 0 | 0 | — |
case-05 | fail→fail | 19,080 | 27,159 | +42% | 1 | 1 | 0% | 2,527 | 5,992 | +137% | 0 | 0 | — |
case-01 | fail→fail | 49,300 | 43,645 | -11% | 1 | 1 | 0% | 8,313 | 9,083 | +9% | 0 | 0 | — |
case-02 | fail→fail | 36,472 | 40,075 | +10% | 1 | 1 | 0% | 6,925 | 8,507 | +23% | 0 | 0 | — |
case-03 | fail→fail | 56,120 | 29,946 | -47% | 1 | 1 | 0% | 8,295 | 7,841 | -5% | 0 | 0 | — |
case-04 | pass→pass | 22,106 | 29,005 | +31% | 1 | 1 | 0% | 2,839 | 5,472 | +93% | 0 | 0 | — |
case-06 | pass→pass | 15,275 | 18,998 | +24% | 1 | 1 | 0% | 2,560 | 4,820 | +88% | 0 | 0 | — |
case-07 | fail→pass | 17,222 | 27,067 | +57% | 1 | 1 | 0% | 3,216 | 5,926 | +84% | 0 | 0 | — |
case-08 | fail→pass | 12,051 | 14,209 | +18% | 1 | 1 | 0% | 1,953 | 3,904 | +100% | 0 | 0 | — |
case-09 | fail→pass | 13,975 | 18,446 | +32% | 1 | 1 | 0% | 2,663 | 5,612 | +111% | 0 | 0 | — |
case-10 | fail→fail | 14,866 | 18,405 | +24% | 1 | 1 | 0% | 2,545 | 4,935 | +94% | 0 | 0 | — |
case-11 | fail→pass | 23,620 | 32,309 | +37% | 1 | 1 | 0% | 4,241 | 7,738 | +82% | 0 | 0 | — |
case-12 | fail→pass | 15,091 | 19,119 | +27% | 1 | 1 | 0% | 2,641 | 5,171 | +96% | 0 | 0 | — |
case-13 | pass→pass | 16,206 | 19,485 | +20% | 1 | 1 | 0% | 2,348 | 5,274 | +125% | 0 | 0 | — |
case-14 | fail→pass | 19,189 | 29,225 | +52% | 1 | 1 | 0% | 3,344 | 6,880 | +106% | 0 | 0 | — |
case-15 | pass→pass | 23,518 | 23,049 | -2% | 1 | 1 | 0% | 2,618 | 5,060 | +93% | 0 | 0 | — |
case-16 | fail→fail | 16,141 | 31,146 | +93% | 1 | 1 | 0% | 2,983 | 7,872 | +164% | 0 | 0 | — |
case-18 | fail→fail | 16,654 | 28,708 | +72% | 1 | 1 | 0% | 2,945 | 6,069 | +106% | 0 | 0 | — |
case-19 | pass→fail | 45,104 | 32,368 | -28% | 1 | 1 | 0% | 7,532 | 6,709 | -11% | 0 | 0 | — |
case-20 | pass→pass | 21,699 | 28,863 | +33% | 1 | 1 | 0% | 4,809 | 7,947 | +65% | 0 | 0 | — |
case-21 | pass→pass | 20,378 | 22,822 | +12% | 1 | 1 | 0% | 4,626 | 7,086 | +53% | 0 | 0 | — |
case-22 | fail→fail | 20,224 | 14,910 | -26% | 1 | 1 | 0% | 2,710 | 4,061 | +50% | 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 +27 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.