AGENT SKILL

agile-7-create-stories

User Stories in Jira from Epic. Triggers: write stories, create user stories, break down epic. After skill 6, before skill 8.

License 许可
MIT
最近更新
2026-11-13
Stars
★ 0
Forks
0

01 前言 Front Matter

name
agile-7-create-stories
description
User Stories in Jira from Epic. Triggers: write stories, create user stories, break down epic. After skill 6, before skill 8.

03 SKILL.md 正文

agile_7_create_stories

Senior Product Manager writing User Stories precise enough for AI dev agents to implement without ambiguity: scan → derive the Story list → interview for gaps → write in Jira → advise.

These Stories feed AI dev agents. Every ambiguity left in an AC becomes a wrong assumption in generated code, so precision here is what saves rework downstream.

Step 1 — Scan existing state

Read the target Epic in Jira in full (summary, description, scope, ACs, Confluence links), follow those links to the PRD (functional requirements for this area) and Specs UI (screens and components for this Epic), and search Jira for Stories already linked to the Epic.

Report the Epic, its goal, and a table of existing Stories (key | summary | status) before doing anything.

  • Epic not in Jira → stop: "I can't find that Epic in Jira. Please run skill 6 first."
  • No Specs UI link → warn, do not stop: "The Epic has no Specs UI link. I'll derive Stories from the PRD functional requirements only, so they may lack screen-level detail. Run skill 3 (INTEGRATE) and re-run this skill to enrich them."

Step 2 — Derive the Story list

Do the analytical work before asking anything.

Map PRD requirements to Stories. Each Story is a single deliverable unit of user value — not a technical task, not a bundle. One user action / one system behaviour is a Story candidate; requirements that can only ship together merge into one Story; requirements that demo independently split into separate ones; and anything that will not fit in one sprint gets split.

Map Specs UI screens to Stories. Each screen or significant state (empty, error, success) typically maps to or contributes to one Story. Cross-reference with the PRD grouping — one screen may cover several FRs, and one FR may span several screens.

Present the derived breakdown and wait for confirmation before writing anything:

| # | Story title (draft) | Covers FR | Covers screen(s) | Est. size |
|---|---------------------|-----------|------------------|-----------|
| 1 | [title] | FR-01, FR-02 | Login screen | S |
| 2 | [title] | FR-03 | Dashboard — empty state, default | M |

Then ask: does the breakdown look right, or should any Stories merge or split? Anything missing that the PRD and Specs UI don't cover? Anything here out of scope for this iteration?

Step 3 — Interview for Story-level gaps

Mandatory per Story: a persona-first title ("As a [persona], I want [action] so that [benefit]") or a clear imperative for technical Stories; a description with context and why it matters; falsifiable acceptance criteria in Given/When/Then; linked Specs UI screen(s); out of scope for this Story; and the Definition of Done. Optionally also: technical notes (known constraints, "must use the existing AuthService", "endpoint exists at POST /api/v1/session"), edge cases the ACs must cover, and dependencies on another Story or system.

Ask when an AC is not falsifiable ("the UI should look good" → "what specifically makes it pass — is there a design reference?"), when the persona is unclear ("user" is not a persona — which one from the PRD?), when a Story touches a system the ADR does not describe ("is this endpoint already built, or does it need creating?"), or when non-trivial edge cases (concurrent edits, pagination, file-size limits) are absent from the Specs UI. Infer and flag the persona in a single-persona Epic, a technical note derivable from the ADR ("ADR specifies JWT — assuming Stories here use the existing middleware"), or the project-wide DoD (flag once, apply to all).

Group questions by Story — never mix questions from different Stories into the same numbered item — and state every assumption in the same message. Never infer silently.

Story 2 — [title]:
1. The AC "user sees the dashboard" is not specific enough — which data, in what format?
2. Is there a loading state, or does data always load instantly?

Story 4 — [title]:
3. Which persona performs this — [Persona A] or [Persona B]?
4. What happens when the file exceeds the size limit — silent fail, error, or redirect?

Assuming across all Stories:
- DoD includes unit tests passing, no lint errors, PR approved by Tech Lead — correct me if yours differs
- All Stories in this Epic use the existing JWT middleware from the ADR

Step 4 — Write Stories in Jira

Issue type Story, linked to the parent Epic. Labels: [project-slug], [epic-slug], [layer]. Status: To Do. Story points: left blank — skill 8 estimates them.

Summary — e.g. As a manager, I want to export the dashboard as PDF so I can share it in reports, or Display empty state on Dashboard when user has no data.

## Context
[1–2 sentences: who this is for, what they are trying to do, why it matters]

## Out of scope for this Story
- [explicit exclusion]

## Technical notes
- [known constraint, existing system, or ADR endpoint reference]

## References
Epic: [PROJ-XXX] · Specs UI screen: [direct Confluence anchor] · PRD: FR-XX

## Acceptance Criteria
- AC1: Given [X], when [Y], then [Z]
- AC2: Given [X], when [Y], then [Z]
- AC3: [edge case] — Given [X], when [Y], then [Z]

## Definition of Done
- [ ] All ACs pass
- [ ] Unit tests written and passing
- [ ] No lint or type errors
- [ ] PR opened and linked to this Story
- [ ] Tech Lead review approved
- [ ] Tested on [staging / local env]
- [ ] No regressions in related Stories
- [ ] UI Story: Specs-UI match **shown, not asserted** — an evidence artefact (a screenshot of the
      screen at each supported viewport) plus an automated a11y check that FAILS the build. See the
      project's QA doc for what counts; "matched" with no artefact is unfalsifiable.

Every AC must be falsifiable (a dev agent can write a test for it), specific (no "fast", "nice", "user-friendly"), and scoped (it does not bleed into another Story's territory).

Step 5 — Resume logic

Re-read every existing Story on the Epic and check each for completeness — title, description, ACs, DoD, Specs UI link. Create only genuinely missing Stories; update only empty or placeholder sections. If the Specs UI changed since a Story was written, comment "Specs UI updated on [date] — AC may need revisiting" rather than editing silently. Never overwrite ACs that have already been through refinement (skill 8).

Step 6 — Advise

✅ Done:
- [N] Stories created under Epic [PROJ-XXX] · [N] already existed and were reviewed
- All Stories linked to Specs UI screens and PRD requirements

⚠️ Still needed (human action required):
- Review Stories with TBD sections: [list]
- Confirm edge cases for: [Stories with possibly incomplete ACs]
- Stories without a Specs UI link: [list — needs design integrated first]

👉 Next step — Skill 8: agile_8_refinement — points, AC validation, final DoD.
   Start with the Stories that have no blocking dependencies.

Principles

  • Analytical first, questions second — derive the Story list from the PRD + Specs UI, then validate it; propose before creating anything in Jira.
  • One Story = one deliverable unit of value — not a task, not a bundle; something demoable on its own.
  • An AC must be falsifiable — a dev agent cannot implement "looks good".
  • A UI Story always carries a direct Specs UI link.

04 同仓库更多 Skills

cedricfarinazzo
cedricfarinazzo/agile-skills

session-audit

Audit the cost and efficiency of Claude Code sessions that ran agile-skills: per-skill and per-subagent spend, a before/after delta normalised by work shipped, and ranked skill improvements. Triggers: audit this session's cost, analyze the session transcripts, compare before and after sessions, is the new version more efficient, where do the tokens go, subagent cost breakdown, find efficiency improvements.

★ 0
cedricfarinazzo
cedricfarinazzo/agile-skills

deep-it-project-review

Autonomously perform a deep, evidence-backed technical review of a software or IT project across applicable architecture, code, security, delivery, operations, and lifecycle domains without modifying project files. Triggers: deep IT project review, technical project review, architecture audit, repository assessment, assess this codebase, review this project.

★ 0
cedricfarinazzo
cedricfarinazzo/agile-skills

test-refactor

Audit one test suite at a time for dead weight, duplication, parallel-unsafety and runtime cost, or characterize behavior before a risky refactor — with production code frozen and good coverage kept, proven per module rather than assumed. Triggers: test refactor, characterize before refactor, characterization tests, clean the tests, refactor the tests, test suite audit, DRY the tests, parallelize the test suite, speed up the tests.

★ 0
cedricfarinazzo
cedricfarinazzo/agile-skills

doc-refactor

Audit every markdown file in a repo — READMEs, docs/, and agent-instruction files — for lies, drift, duplication and bloat, then ticket and ship the cleanup as a sequenced PR train, with source code frozen and every surviving claim verified against the repo rather than read. Triggers: doc refactor, refactor the docs, clean the docs, documentation audit, README audit, docs are stale, DRY the docs, audit every markdown file.

★ 0
cedricfarinazzo
cedricfarinazzo/agile-skills

deep-refactor

Audit an entire codebase for defects, waste and debt, then ticket and ship the findings as a sequenced PR train — without breaking the test contract. Triggers: deep refactor, refactor audit, clean the codebase, DRY pass, audit and fix everything.

★ 0
cedricfarinazzo
cedricfarinazzo/agile-skills

agile-sprint-drain

Drain the active sprint to a fixed point: auto-alternate agile-10-implement (build) and agile-11-merge-train (merge) until both empty (DRAINED) or blocked (STUCK). Optional concurrency=N, dispatch=session, max-merges=N. Triggers: drain the sprint, run the sprint to completion, implement and merge until done, clear the whole board, ship the sprint.

★ 0