Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Idea to PRD: An Honest Draft With Every Assumption Flagged

    1

    A short clarifying interview turns a one-liner into a structured PRD with per-section confidence scores. It never invents market data or sources.

    Free

    1 useSecurity scanned
    Idea to PRD: An Honest Draft With Every Assumption Flagged

    Idea to PRD: An Honest Draft With Every Assumption Flagged

    Example session with this skill installed

    Turn my idea into a PRD: a retro tool where engineers can raise issues anonymously and the action items actually get done.

    • Read your context and instructions
    • Compiled the idea to prd

    Idea to PRD first runs a short clarifying interview (users, problem, competitors, business model, scope, constraints, success criteria). From the answers alone, with no research and nothing invented, it drafts:

    Draft PRD — forge-pulse-fast-feedback-team-heartbeat (v1)

    Problem (confidence 55%)

    Engineering teams run retrospectives but

    most action items die before the next sprint. The founder believes the root causes are (1) low psychological safety, so real issues go unspoken, and (2) no durable follow-through on the actions that do get raised.

    • Pain is felt most by individual contributors who don't feel safe raising problems.
    • Existing retro tools focus on the meeting, not on what happens after.

    ⚠️ ASSUMPTION: root-cause claims come from the founder's intuition, not validated user research.

    User Personas (confidence 50%)

    • IC engineer (primary): wants to raise real concerns without fear of attribution.
    • Team lead / facilitator: runs the retro and is accountable for follow-through.

    ⚠️ ASSUMPTION: persona pain points are inferred from the interview, not from interviews with real target users.

    Scope & Non-Goals (confidence 60%)

    In scope (MVP): anonymous feedback capture, simple upvoting, action items with owners.

    Non-goals: multiple retro formats, cross-team analytics, third-party integrations.

    ⚠️ ASSUMPTION: credible anonymity can be delivered in the MVP — not yet validated technically.

    Success Metrics (confidence 40%)

    • Action completion rate rises from an assumed baseline of ~40-50% to 70%+ within 3 months of adoption.
    • Participation: a majority of team members submit at least one note per retro.

    ⚠️ VAGUE_METRIC: baseline completion rate is assumed, not measured.
    ⚠️ UNKNOWN: no instrumentation plan for measuring completion.

    User Stories (confidence 65%)

    1. As an IC, I can submit a retro note anonymously so I can be honest.
    2. As a team, we can upvote notes so the most important themes rise.
    3. As a facilitator, I can turn a theme into an action item with an owner and due date.

    Technical Considerations (confidence 45%)

    • Real-time collaboration during the retro implies WebSockets.

    Anonymity is the hard part: the data model must not store authorship in a way a facilitator or admin could reverse.

    • Duplicate-vote prevention without identity is an open problem.

    ⚠️ ASSUMPTION: WebSockets are required — not yet validated against actual usage patterns.
    ⚠️ UNKNOWN: anonymous duplicate-vote prevention approach is undecided.

    Risks & Open Questions (confidence 50%)

    • The product may be

    solving a culture problem with tooling — tracking actions more visibly may not make teams complete them.

    • Free, established retro tools may be "good enough," making willingness-to-pay unclear.
    • Open question: who is the buyer, and is this a standalone product or a feature?

    ⚠️ RISK: tooling may not fix an organizational/culture problem.
    ⚠️ UNKNOWN: willingness to pay and the buyer are undetermined.

    Explicit assumptions

    • The core problem is low psychological safety plus poor follow-through (founder's hypothesis, unvalidated).
    • Teams will trust an anonymity guarantee enough to speak freely.
    • A standalone tool is worth adopting over existing free retro tools.

    Contradictions flagged

    • Notes are meant to be permanently anonymous, yet action items have named owners — an owner may be inferable from the note that spawned the action.

    Next: it points you to the lowest-confidence sections (Success Metrics at 40%, Technical Considerations at 45%) and the flagged assumptions as the things to validate first, and offers to red-team the draft.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most AI-written PRDs read confidently and quietly invent things: market sizes, competitors, statistics, sources. Idea to PRD drafts only from what you actually told it, and flags everything else.

    How it works

    1. Give it an idea. A one-liner, a rough note, or a paragraph.
    2. Clarifying interview. Up to three rounds of 4–8 questions covering target users, the core problem and how painful it is, competitors, business model, scope and timeline, technical constraints, success criteria, and what makes it different. Each round sharpens the next.
    3. Honest draft. A structured PRD built only from your answers: empty citations, assumptions flagged, and a confidence score on every section.
    4. Validated output. The PRD is checked against a strict schema, then rendered as readable markdown.

    What you get

    Seven sections: problem, personas, scope and non-goals, success metrics, user stories, technical considerations, risks and open questions. Plus:

    • ⚠️ inline flags for every assumption, vague metric, unknown and risk
    • An explicit assumptions list: what to validate first
    • Contradictions found in your own answers
    • Per-section confidence that reflects how much you told it versus how much it assumed

    What's inside

    • SKILL.md: the interview and drafting flow
    • references/synthesizer.md: the PRD-writer prompt and schema
    • scripts/reckoning.mjs: a short, readable validator and renderer with no dependencies (Node 18+)
    • examples/expected-prd.md: a real draft PRD

    Known limitations

    • No web research. Citations stay empty by design; the draft is as good as your answers.
    • Rendering needs Node 18+. Without it, the agent formats the output by hand.
    • Want the draft stress-tested too? Use Reckon (full flow) or Spec Red-Team on the result.

    Free and open source (MIT): github.com/alchemyguy/reckoning

    No web research: citations stay empty by design, so the draft is only as good as your interview answers. The interview runs at most three rounds. Automatic validation and rendering need Node.js 18+. It drafts a PRD but doesn't critique it; use Reckon or Spec Red-Team for that.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 29 days ago

    • Free with an Agensi account

    Requires

    Node.js

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    File Scopes

    .reckoning/**
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    A plain prompt fills gaps with confident-sounding inventions: market sizes, competitors, statistics. Idea to PRD interviews you first, drafts only from your answers, flags every assumption inline, and scores each section's confidence by how much you actually told it.

    Target users, the core problem and how painful it is, the competitive landscape, the business model, scope and timeline, technical constraints, success criteria, and what makes the idea different. It asks 4–8 questions per round for up to three rounds, getting sharper each time, and stops early once it has enough.

    Seven sections (problem, personas, scope and non-goals, success metrics, user stories, technical considerations, risks and open questions), each with a confidence score and inline flags for assumptions, vague metrics, unknowns and risks. It ends with an explicit assumptions list and any contradictions found in your answers.

    No. It runs inside your own agent session. The bundled script makes no network calls and needs no npm install; it only reads and writes JSON files in a local .reckoning/ folder.

    Any agent that supports the SKILL.md standard, including Claude Code, Cursor, Codex CLI, GitHub Copilot and Gemini CLI. It uses a structured question tool for the interview when the agent has one, and plain chat otherwise.

    New in Business & Operations