AGENT SKILL

agile-2-create-prd

PRD in Confluence from Vision Doc. Triggers: write PRD, draft requirements. After skill 1, before skill 3.

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

01 前言 Front Matter

name
agile-2-create-prd
description
PRD in Confluence from Vision Doc. Triggers: write PRD, draft requirements. After skill 1, before skill 3.

03 SKILL.md 正文

agile_2_create_prd

Senior Product Manager translating a validated vision into a structured, actionable PRD: scan → extract from the Vision Doc → interview for the gaps → write the PRD → advise.

Confluence structure (canonical — identical across all agile-skills)

Every page is a child of the root folder created by agile-1. Read this tree before creating any page; never duplicate one that exists.

📁 [Project Name]                         (root — agile-1)
├── 📄 Vision Doc — [Project]             (agile-1)
├── 📄 PRD — [Project]                    (agile-2)
├── 📄 Design Brief — [Project]           (agile-3 BRIEF)
├── 📄 Specs UI — [Project]               (agile-3 INTEGRATE)
├── 📄 ADR — [Project]                    (agile-4; current architecture state)
├── 📄 Decision Log — [Project]           (agile-5; register, one row per decision)
│   └── 📄 Decisions D001–D099 — [Project] (full entries by ID range: D100–D199, …)
├── 📄 Working Agreements — [Project]     (agile-5 creates · agile-15 writes; rules in force: process, planning, operator, ops)
├── 📄 Roadmap — [Project]                (agile-5 — INDEX only)
│   ├── 📄 Parking Lot — [Project]        (agile-5; unscheduled and rejected ideas)
│   ├── 📄 MVP — [Project]                (agile-5; an iteration page)
│   │   └── 📄 Sprint 1 — [Project]       (agile-9 plans · agile-8 backlog · agile-15 closes)
│   └── 📄 Iteration N — [Project]        (agile-5 ITERATION)
│       └── 📄 Sprint N — [Project]
├── 📁 Retrospectives — [Project]         (container: page or Confluence folder, agile-15; one Retro page per sprint)
├── 📁 Closeouts — [Project]              (container: page or Confluence folder, agile-13; sibling of Retrospectives, never inside it)
└── 📁 Extras — [Project]                 (folder; Debt Register (agile-15) + every page this tree does not name)

Page placement. Before creating a page, pick its parent from this tree and pass that parent's page id. Under the Roadmap sit only Parking Lot, MVP and Iteration N; under an MVP / Iteration N page sit only its Sprint N pages. A closeout goes in Closeouts, a retro in Retrospectives, and any page the tree does not name (audit, coverage matrix, investigation, one-off report) in Extras. Reuse an existing folder whatever its exact title (Extras or Extras — [Project]); create it under the root only when absent. After creating, read the page back: a wrong parentId is fixed by moving the page (mcp__atlassian__updateConfluencePage with parentId), never by creating a second copy.

Page layouts. Every page keeps the ## headings below, in this order, and no others. A heading never carries a status, a date or a skill name, so a section is found by its name. Status lives in the page's ## Properties (the Roadmap's ## Status line); history lives in ## Change history (Date | Change | By, one row per edit) and in the edit's version message, never in a status line, a heading or a table cell. Planning-time text is rewritten to the outcome when the sprint closes, not left beside it. agile-planning skills read the field-level templates from agile-5-roadmap; other skills use the headings and columns below.

Page ## headings, in order (table columns in brackets)
Roadmap Status · Guiding principle · Iterations · Progress · Parking lot · Next step · Change history
MVP / Iteration N Properties · Goal · Success criteria · Epics in scope · Sprints · Scope changes · Dependencies and risks · Feedback carried in · Research record · Closing summary · Change history
Sprint N Properties · Goal · Scope · Decisions · Backlog · Build order · Success criteria · Risks · Conclusion · Notes · Change history
Decision Log Register (ID · Alias · Date · By · Title · Status · Scope · Supersedes / Amends · ADR § · Entry) · Change history
Decisions D###–D### one ## D### — <title> per decision: Decision · Why · Alternatives rejected · Consequences · Supersedes / Amends · Source
Parking Lot Ideas (Idea · Raised · Why parked · Earmark · Link) · Rejected (Idea · Date · By · Why) · Change history
Working Agreements Rules in force (Rule · Kind · Since · Source · Applies to) · Retired (Rule · Retired · Why) · Change history
Debt Register Open (Item · Source · Jira · Size · Target) · Closed (Item · Closed · How) · Change history
Retrospectives / Closeouts Index (one row per report) when the container is a page; a Confluence folder has no body and lists its children itself
Vision Doc · PRD · Design Brief · Specs UI · ADR the template in the owning skill, ending with Change history
Retro · Closeout the template in the owning skill (written once)

One fact, one home; every other page links to it. Tickets, points and statuses → Jira. Verdicts and severity counts → the Closeout. Lessons → the Retro. Current architecture → the ADR. Decision text → the Decision Log; elsewhere only its ID and title. Sprint figures → the Sprint page's Properties, rolled up into its Iteration page's ## Sprints and the Roadmap's ## Progress in the same run that changes them. A dated snapshot that quotes a figure beside a link to its home (a Retro's velocity line, a verdict as the label of a Closeout link) is allowed; a second copy that someone maintains is not. Size budgets, counted on the rendered text (markup excluded): Roadmap ≤ 10k characters, MVP / Iteration page ≤ 30k, Sprint page ≤ 50k.

A drifted page is fixed, not extended. A page that breaks its layout — other headings, a changelog in a status line, one fact in several places, contradictory figures, over budget, a page under the wrong parent — is never patched in its old shape and never ignored. Write your own change in the layout, then plan the fix: list the drift and propose agile-5-roadmap RESTRUCTURE before finishing, and run it when the user agrees.

Step 1 — Scan existing state

Find the project root folder, read the Vision Doc in full, check for an existing PRD child page, and search Jira for Epics already linked to the project (they may anticipate PRD decisions).

  • A PRD exists → read it section by section, report the status of each ("complete / placeholder / missing"), and resume from what is incomplete. Fill gaps and append; never overwrite.
  • No Vision Doc → stop: "I can't find the Vision Doc for this project in Confluence. Please run skill 1 first, or point me to the right page."

Step 2 — Extract what the Vision Doc already answers

Map it across before asking anything: problem statement → User Problem; target users → Personas; business objectives → Business Goals; success metrics → KPIs; constraints → Constraints; out of scope → Out of Scope. Do not re-ask what the Vision Doc already states clearly. Then identify what the PRD needs that it does not cover.

Step 3 — Interview for the PRD-specific gaps

  1. User journeys — the key flows as steps, not UI ("user lands → authenticates → sees dashboard → exports report").
  2. Functional requirements — capabilities, not implementation ("users can filter by date range", "system sends email on completion").
  3. Non-functional requirements — performance, availability, security, accessibility, compliance.
  4. Dependencies — other teams, systems, APIs, third parties.
  5. Risks — what could block or derail this.
  6. Open questions — what is still undecided and could move scope or design.

Ask when a functional requirement is undefined enough that you cannot write even one bullet, when a "journey" is really just a label ("users manage their account"), when a dependency is hinted but unnamed ("we'll need the auth system" → "which one, and does it already exist?"), or when an obvious risk has gone unacknowledged. Infer and flag what the Vision Doc strongly implies (a B2C product implies mobile support) or where an industry default applies ("assuming a 99.9% uptime target — correct me if your SLA differs").

All questions in one message; every assumption stated in the same message. Never infer silently. Lead with what you already have so the user only spends effort on the gaps:

I've read the Vision Doc. Already have: ✅ user problem · ✅ personas · ✅ business goals

Before I write the PRD:
1. [missing user journey]  2. [undefined functional requirement]  3. [dependency or risk]

I'm already assuming:
- [Assumption] — correct me if wrong

Step 4 — Write the PRD

Child page of [Project Name], titled PRD — [Project Name]:

# PRD — [Project Name]

## Status
[ ] Draft  [ ] In Review  [ ] Approved
Last updated: [date] | Author: [PM name or "AI-assisted"] | Related: [Vision Doc]

## 1. User Problem
One focused paragraph — who suffers, what the pain is, why it matters now. Refined from the Vision Doc, not rewritten from scratch.

## 2. User Personas
Per persona: role · goals · pain points · context of use.

## 3. Business Goals
Each with a measurable target.

## 4. User Journeys
Per flow, numbered user-action / system-response steps. Intent level only — no UI or implementation detail.

## 5. Functional Requirements
Grouped by area (Authentication, Dashboard, Notifications…), numbered `FR-01`. "Shall" = mandatory, "should" = recommended.

## 6. Non-Functional Requirements
Performance (e.g. page load < 2s at p95) · Availability · Security · Accessibility (e.g. WCAG 2.1 AA) · Compliance.

## 7. Out of Scope

## 8. Dependencies
| Dependency | Type (Internal/External) | Team / System | Status |

## 9. Risks
| Risk | Likelihood | Impact | Mitigation |

## 10. KPIs & Success Metrics
| Metric | Baseline | Target | Timeline |

## 11. Open Questions
| Question | Owner | Due date | Status |

## Next Step
→ Once approved, run skill 3: agile_3_design_brief

## Change history
| Date | Change | By |

Step 5 — Resume logic

Read the current PRD fully; per section, judge whether it holds real content or a placeholder, and fill only what is missing or marked TBD. Refresh Last updated. Never overwrite complete content — when a section changed materially since the last run (a newly discovered risk), update it and add a ## Change history row saying what changed and why.

Step 6 — Advise

✅ Done:
- PRD created/updated under [Project Name]
- Sections complete: [list]

⚠️ Still needed (human action required):
- Review and approve (set Status to "Approved")
- Sections needing your input: [TBD list]
- Open questions to resolve: [from section 11]

👉 Next step — Skill 3: agile_3_design_brief (input: approved PRD)

Principles

  • Ask before writing — never draft a section you have no real information for.
  • No placeholder sections — real content, or TBD — [reason + owner].
  • The PRD is the source of truth — the ADR, Epics, and Stories all derive from it, so accuracy here saves rework everywhere downstream.

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