agentleFS
shawnclyborshawnclybor/clybor-claude-tooling/.claude/skills/prd-writer/SKILL.md

prd-writer

Skill6 starsChanged 29 days agoMITView on GitHub

Write a PRD (product requirements doc) for a new feature or non-trivial change. Output is a markdown file with problem statement, binary success criteria, in/out of scope, dependencies, and risks. Use at the start of any work that warrants a quality gate before implementation. Triggers — "write a PRD", "draft a PRD for X", "start with a PRD", "PRD for [feature]", "spec out [feature]". The success criteria written here become the verification gates downstream — they must be binary (yes/no answerable), not vibes.

40 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use experiment-designer, user-research and 2 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • Dokkabei97Skill

    experiment-designer

    Designs A/B tests and product experiments with hypothesis structuring, sample size calculation, duration estimation, and result interpretation. Unified framework for both product and marketing experiments.

    Also in 3 projects that use this

  • AnthropicSkill

    user-research

    Plan, conduct, and synthesize user research. Trigger with "user research plan", "interview guide", "usability test", "survey design", "research questions", or when the user needs help with any aspect of understanding their users through research.

    Also in 3 projects that use this

  • yannickYamoSkill

    product-strategy

    Build product strategy that drives real choices. Use when defining where to play and how to win, creating strategy canvases, defining ICP/Anti-ICP, positioning products, setting pricing strategy, planning GTM motions, or structuring strategic bets. Part of the Modern Product Operating Model collection.

    Also in 3 projects that use this

  • OpenAISkill

    brainstorming

    You MUST use this before any creative work - creating features, building components, adding functionality, or modifying behavior. Explores user intent, requirements and design before implementation.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 4 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 28 edited versions we read keep these.

  • 4

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    If problem_statement is missing, ask before drafting.

    line 29
  2. Distinctive:

    If the problem statement references external claims, libraries, papers, or product names, invoke:

    line 33
  3. Very specific:

    If a criterion cannot be made binary, it belongs in ## 7.

    line 63
  4. Distinctive:

    Before writing the file:

    line 67
  5. Distinctive:

    Do I have the problem statement from the user, in their words?

    line 69
  6. Very specific:

    If the PRD cites external claims, has evidence-auditor verified them?

    line 75
  7. Distinctive:

    If the PRD is wrong, all three are wrong.

    line 85

“Common” is measured: kept by at least 6 in 10 of the 28 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

PRD Writer

A PRD here is short and operational. It is the contract between the problem and the implementation. If the success criteria are wrong, every downstream stage is wrong.

When to use

  • Starting a new feature with more than a trivial change
  • A refactor that crosses module boundaries
  • An integration where the contract with an external system matters
  • Any change where the user wants a quality gate before code lands

When NOT to use

  • One-line fixes, typos, dependency bumps
  • Spike work where the goal is to learn
  • Bug fixes where the fix is mechanically obvious

Inputs

  • feature_slug — kebab-case identifier
  • problem_statement — one paragraph: what is being built and why
  • (optional) source_material — links, screenshots, prior art

If problem_statement is missing, ask before drafting. Do not infer.

Source-gathering (when needed)

If the problem statement references external claims, libraries, papers, or product names, invoke:

  • research-analyst — for deep multi-source synthesis (3+ sources to compare or summarize)
  • search-specialist — for a single precise lookup (API signature, version, exact quote)
  • evidence-auditor — to verify any claim the draft PRD attributes to an external source

These run before drafting, not after. A PRD with unverified citations is worse than one with no citations.

Output

docs/PRDs/<feature_slug>.md, copied from templates/prd-template.md and filled in. Sections:

  1. Problem — 1-3 paragraphs of current state + pain
  2. Success criteria — binary checkboxes
  3. In scope — what is included
  4. Out of scope — anti-scope
  5. Dependencies — what must land first
  6. Risks — specific failure modes with likelihood and mitigation
  7. Open questions — unresolved items

The success criteria rule (most important)

Criteria must be binary. Each one passes a single test:

Can a fresh reader, given the criterion plus the implementation, answer yes or no without judgment?

Bad (vibes): "Improve performance", "Make the code cleaner", "Handle edge cases gracefully"

Good (binary): "P95 latency on /login under 200ms across 100 test requests", "Cyclomatic complexity ≤ 8 for every function in auth/", "On invalid email, the API returns HTTP 400 with {error: 'invalid_email'} (verified by integration test)"

If a criterion cannot be made binary, it belongs in ## 7. Open questions until resolved.

Pre-flight checklist

Before writing the file:

  1. Do I have the problem statement from the user, in their words?
  2. Have I distinguished problem from solution? (PRD is about WHAT, not HOW.)
  3. Is every success criterion binary? If not, fix or move to open questions.
  4. Is there a clear out-of-scope section?
  5. Have I listed dependencies even if "none"?
  6. Are risks specific failure modes, not generic phrases?
  7. If the PRD cites external claims, has evidence-auditor verified them?

What happens next

The PRD feeds three downstream skills:

  • task-plan — decomposes the PRD into a checkbox task list
  • verify — uses the success criteria as the verification gates
  • quality-review — runs the 3-agent team against the PRD before implementation starts

If the PRD is wrong, all three are wrong. Spend the time here.

Anti-patterns

  • Inferring success criteria from vibes — ask the user
  • Burying solution in the problem — PRD describes WHAT, not HOW
  • Vague risks — "things might break" is not a risk; "the third-party API rate-limits at 100 req/min" is
  • Skipping out-of-scope — if you do not write the anti-scope, scope creeps
  • 500-word PRDs for 20-line features — the PRD should be proportional to the work
  • Unverified citations — every external claim runs through evidence-auditor before the PRD ships
  • metadata-fetcher — invoked when the PRD references a paper, library, package, or repo by identifier (DOI, package name, repo URL). Returns the structured record (version, license, OA status, last release) so the PRD cites accurate metadata. Mechanical lookup; no analysis.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.