Prd Authoring

prd-authoring

> Anchor: A PRD is a hypothesis, not a spec. Reduce uncertainty, not scope.

When to use

Use this skill when the user needs to write or restructure a product requirements document for a feature, surface, or initiative. Works for:

• A new feature that needs alignment before build.
• An existing feature that needs a retroactive PRD for handoff or audit.
• A discovery-stage concept that needs to be reduced to a testable hypothesis.

Do not use when:

• The user is proposing a feature pre-scoping. Use feature-proposal instead.
• The user is scoping delivery work (time, phases, commercial). Use proposal-scoping.
• The user is triaging existing tickets. Use ticket-triage.

Philosophy

A PRD is a hypothesis under uncertainty. It is a bet, placed in writing, with kill criteria and instrumentation that will tell you whether the bet paid off. Treating it as a contract for engineering is a category error. Treating it as a creative document is a cost the team pays for months.

Five concepts from the handbook apply directly. Output vs outcomes: write what should change in user behavior or system state, not what the team will produce. Product as a portfolio of bets: the PRD names the bet and its size. Reduce uncertainty, not scope: the shortest PRD is not the best; the PRD that kills the riskiest unknown first is. PM as anthropologist: the problem statement comes from observation, not invention. Discovery is a practice: the PRD carries evidence of what has already been validated and what has not.

Voice rules (apply to all outputs)

• No buzzwords: leverage, empower, synergy, resonate, tapestry, delve, elevate, captivate, explore, dynamic, testament.
• No em dashes.
• No filler openings.
• Decision-oriented over descriptive.
• High-context, no overexplaining.
• Standard sentence case for formal outputs.
• ES outputs match the register of the handbook: direct, self-aware, no reverence for process, comfortable naming failure.
• EN outputs use direct US business register. Do not soften bluntness into corporate-speak.
• Language switches based on audience, not author.

Bilingual handling

• ES. Rioplatense register. Vos over tú. Keep technical terms in English where that is standard in industry use: PRD, backlog, sprint, stakeholder, rollout, feature flag, AB test, KPI. Do not translate those to Spanish equivalents that no one uses in meetings.
• EN. Direct US business register. Short sentences. No British spelling unless the target audience is UK or AU.
• Choose the language of the audience that will read the document, not the language the user types in. Ask once if ambiguous.

Method

Step 1. Classify the feature by maturity

Pick one. This decides which template to use.

| Maturity | Signals | Template | |---|---|---| | Discovery | Problem is not yet confirmed. No user interviews. No data. | prd-lean | | Shaping | Problem confirmed. Solution space open. Some data. | prd-lean | | Build | Solution committed. Dependencies mapped. Timeline real. | prd-full | | Post-hoc | Feature shipped. PRD written for handoff or audit. | prd-full |

When in doubt, use prd-lean. Full PRDs written too early are expensive fiction.

Step 2. Write the problem before anything else

Do not open the document with scope. Open it with the problem.

The problem statement answers four questions:

1. Who feels it. A specific user type, not "users" or "customers".
2. When they feel it. A moment in the product, a frequency, or a trigger.
3. What they do today. The current workaround, detour, or resignation.
4. What that costs. Time, revenue, churn, support load, or a strategic gap.

If you cannot answer all four, the PRD is not ready. Flag the gaps instead of guessing.

Step 3. State the hypothesis

Format:

> If we ship X for user Y, we expect metric Z to move by roughly W, because reason R.

This forces scope, audience, metric, magnitude, and causal logic into one sentence. If the sentence is vague, the PRD is vague.

Step 4. Define scope by exclusion

Write three lists:

• In. What we are committing to build now.
• Out. What we are choosing not to build now, and why.
• Not ever. What we believe is wrong for this feature, not just deprioritized.

The Out list is the most important. It is where teams lose months.

Step 5. Write acceptance criteria in user-observable terms

Each criterion is something a user or another system could verify without access to the codebase. "User receives an email within 5 minutes of event X" is a criterion. "The queue retries twice on failure" is a system note, not an acceptance criterion.

Step 6. Name the instrumentation before the build

Before the feature ships, define:

• Which events fire.
• Which property shape they carry.
• Which funnel they roll up into.
• Which dashboard they populate.
• Who owns that dashboard.

If the PRD does not name these, you will not know whether the bet paid off. Cross-reference analytics-instrumentation if the taxonomy does not yet exist.

Step 7. Name the kill criteria

A PRD without kill criteria is a one-way door. Write:

• What result or signal would cause us to roll back.
• What result or signal would cause us to pause and rethink.
• How long we wait before checking.

Step 8. List the open questions

Every non-trivial PRD has open questions. Write them down with an owner and a deadline. Do not bury them in slack threads.

Step 9. Read the PRD as a stranger

Before handing off, read it as someone who was not in the conversation. If a sentence only makes sense with prior context, fix it or move it to a linked appendix.

Templates

• templates/en/prd-lean.md: discovery and shaping stage.
• templates/en/prd-full.md: build and post-hoc stage.
• templates/es/prd-lean.md: etapa de discovery y shaping.
• templates/es/prd-full.md: etapa de build y post-hoc.

Related skills

• feature-proposal: use it first when the feature is a pre-scoping bet, not a committed build.
• proposal-scoping: use it when scoping delivery work (time, phases, commercial), not product behavior.
• analytics-instrumentation: cross-reference it to build the instrumentation section when the taxonomy does not yet exist.
• ticket-triage: use it when the work is triaging existing tickets rather than writing a spec.

Anti-patterns

This skill must never produce:

• A PRD that opens with a solution or a screen.
• A PRD that lists features without naming the problem they solve.
• A PRD that has no instrumentation section.
• A PRD that uses "users" or "customers" as the only audience descriptor.
• A PRD that treats engineering estimates as commitments before scope is settled.
• A PRD that hides open questions inside meeting notes instead of surfacing them.
• A PRD written in the style of a press release when the audience is internal.
• A PRD with acceptance criteria that require reading the codebase.
• A PRD that mixes English and Spanish inconsistently within a single section.
• A PRD that cites "industry best practice" without a concrete internal reason.

Checklist before delivery

• [ ] Problem statement answers the four questions in step 2.
• [ ] Hypothesis is a single sentence in the step 3 format.
• [ ] Out list has at least three items.
• [ ] Acceptance criteria are user-observable.
• [ ] Instrumentation section names events, properties, and an owner.
• [ ] Kill criteria are concrete and time-bounded.
• [ ] Open questions each have an owner and a deadline.
• [ ] No buzzwords from the voice rules appear anywhere.
• [ ] Document reads cleanly to a stranger who was not in prior meetings.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: federicoimparatta
• Source: federicoimparatta/operator-skills
• License: MIT

Install and usage instructions live in the source repository linked above.