Define the first release and leave room for good implementation

Qiaomu AI PRD

From an Idea to a Developer Handoff

Qiaomu’s specification skill turns a short product idea into a summary card, an 11-chapter PRD and observable acceptance scenarios. It helps define what belongs in the first release, how data moves and how failures recover, so a developer or coding assistant can work from explicit decisions.

By 向阳乔木

Setup & first taskView on GitHub
Qiaomu AI PRD: From an Idea to a Developer Handoff project image

Workflow and outputs

  • Assistants with local skill support
  • Markdown PRD
  • Optional Python 3 checker
GitHub stars
213
Version
0.2.0
Checked
Categories
Product & ProjectWriting & Research

Specify one complete user journey first.

Use concise mode to detail P0 while keeping later scope brief. Review failure paths, persistence and acceptance evidence before adding depth or frontend/backend perspectives.

Pick a task and get started

After installation, name qiaomu-ai-prd in your request and provide an idea or an existing PRD. The tasks cover drafting, handoff revision and document review. Replace the sample paths with actual files.

What do you need to deliver?

Start from an idea and establish the smallest useful scope. Provide the audience, platform and requirements that cannot change; clarify choices that would materially change the product.

  1. Install the complete skill

    Run the author’s command in a local terminal and choose the client and scope. It installs instructions and supporting files; use them in an assistant session that can access your project material.

    npx skills add joeseesun/qiaomu-ai-prd
    Read the author’s setup guide

    Confirm the reference files are available

    Ask the assistant to locate SKILL.md, the three references and scripts/lint_prd.py. Use the actual installation path; do not assume a global ~/.agents/skills location. Saving needs a writable workspace; the checker additionally needs Python 3.

    Read the installation and verification notes

  2. Give your AI this first task

    These are editorial usage examples, not tested skill runs or generated PRDs. The skill defaults to Chinese; request another language explicitly. Its checker looks for Chinese headings, so English documents need a separate review.

    Use qiaomu-ai-prd [精简模式] to draft a PRD for a personal reading list. Users manually add a title and URL, categorize entries, mark them read and export their list. P0 is a single-user web tool without article scraping, accounts or cloud sync. Write in English and keep all 11 chapters plus one opening summary card. Separate hard constraints, recommended defaults and creative latitude. Specify empty lists, invalid URLs, duplicates, failed saves and recovery; include a versioned data model and export format. Treat numeric performance figures as proposed targets with measurement conditions. Add 3–5 acceptance scenarios specifying the evidence to collect after implementation, and preserve honest unknowns. Save docs/reading-list-prd.md. Explain whether the bundled Chinese-heading checker can validate this version; record any checks you could not run. Do not implement the app.

  3. Check what you receive

    • A saved 11-chapter PRD with detailed P0 behavior and brief P1–P3 scope.
    • Module diagrams, states and recovery paths that agree with the persistence and export model.
    • 3–5 acceptance scenarios with environment, actions, expected outcomes and evidence; document-check status is explicit.
    Read the author’s notes for this task
More about this skill

qiaomu-ai-prd makes product decisions explicit. It infers a conservative initial scope from a short brief, explains its defaults and asks about ambiguities that materially affect the platform, budget, data ownership or scope. Its constraint categories help preserve essential behavior while allowing implementation choices to improve.

The output combines one opening summary card with 11 chapters. Concise mode reduces detail in later tiers; deep mode expands edge cases. Diagrams, states, data, architectural decisions and acceptance scenarios should describe the same product rather than merely fill headings.

A bundled Python checker can catch structural omissions. The author also requires honest uncertainty, so this guide distinguishes document checks, product judgment and acceptance testing after implementation.

Sources and verification

Sources checked

Reviewed the README, full skill, three references and checker source. Atlasnote has not installed or run the skill or the third-party checker.

Author documentationOriginal skillDocument quality reviewChecker source

Choose the next step

Use PM Skills to organize user evidence. Once requirements are clear, Addy’s engineering skills can guide specifications, plans and implementation.

PM SkillsProduct-management plugins for interview synthesis, assumption tests, PRDs and roadmaps.Addy Osmani Agent SkillsAddy Osmani’s workflows organize AI coding around specs, small tasks, tests and review.Career OpsWeigh postings against your evidence, prepare CVs and answers, rehearse interviews and track the search.Qiaomu GoalTurn a vague request into outcomes, checks, constraints, boundaries and completion conditions.