- Home
- Skills
- Business & Operations
- Spec-Driven PRD Writer
More screenshots
Works with the AI Agents you already use
Spec-Driven PRD Writer
Most PRDs an agent writes are a wall of prose: no feature IDs, no order to build in, and criteria nobody can check.
Free
Spec-Driven PRD Writer
Example session with this skill installed
Write a PRD for a small web app where a book club proposes books, votes on the next one and records meeting notes. Save it in docs/.
- Read your context and instructions
- Compiled the spec-driven prd writer
It asks first. One question at a time, starting with the one most decisions depend on:
Question 1: Is the app for one book club, or for any number of clubs?
- A. One club. Your club only, with one shared list of members.
B. Many clubs. Anyone can create a club and invite members, and each club's books, votes and notes are kept separate.
- C. Something else. Tell me how you see it.
The user answered "Use sensible defaults for the rest, list them, and go ahead", so it listed every default it chose (one club of up to 50 members, invitation-only joining, approval voting with secret ballots, meetings with a place or a link) and wrote the PRD.
The PRD: docs/book-club-hub-prd.md, "Book Club Hub", in nine sections: Executive Summary, Problem and Opportunity, Target Audience, Objectives, User Stories, Functionalities, Out of Scope, Dependency Graph, Acceptance Criteria.
| # | Feature | Priority | Dependencies |
|---|---|---|---|
| F01 | Member Accounts and Sign-In | 1 | None |
| F02 | Membership and Invitations | 1 | F01 |
| F03 | Book Proposals | 1 | F02 |
| F04 | Voting Rounds | 1 | F02, F03 |
| F05 | Meeting Scheduling | 1 | F04 |
| F06 | Meeting Notes | 1 | F05 |
| F08 | Email Notifications | 3 | F02, F04, F05 |
| F09 | Book Ratings | 2 | F05 |
| F07 | Club Home and Reading History | 2 | F04, F05, F06, F09 |
Build order (execution waves): 1: F01 (foundation) · 2: F02 · 3: F03 · 4: F04 · 5: F05 · 6: F06, F09, F08 · 7: F07.
Acceptance criteria you can check, for example for F04, Voting Rounds:
- A round opens with 2 to 12 candidates; 1 or 13 candidates are refused
- With 5 candidates a member can pick up to 3 books; with 2 candidates, only 1
- A tie leaves the round "Tie — waiting for an organizer" until an organizer chooses one of the tied books
and cross-feature checks such as "The turnout denominator in an open round (F04) equals the number of active members (F02)".
A follow-up, "Add a feature: members rate a book after its meeting", added F09 and updated the dependency table, the waves and the criteria in place.
The run ended, once, with a short note in English: "Your PRD is ready. Next: turn it into a finished, verified project", pointing to the Spec-Driven Dev Pack, which reads this PRD as it is. The note is not written into the PRD.
Connects securely to your tools. The creator never sees your data.
About this skill
Most PRDs an agent writes are a wall of prose: no feature IDs, no order to build in, and criteria nobody can check. Spec-Driven PRD Writer asks what it cannot infer, then writes a PRD in nine fixed sections that a coding agent can build from, feature by feature.
Free. The whole skill, with nothing held back.
What you get
A PRD saved in your project, in nine sections: Executive Summary, Problem and Opportunity, Target Audience, Objectives, User Stories, Functionalities, Out of Scope, Dependency Graph and Acceptance Criteria.
- Every feature has an ID (F01, F02…) used in its user stories, functionalities, dependencies and criteria.
- Dependencies and build order. Section 8 lists what each feature depends on and works out the execution waves, with the foundation features first, plus a Mermaid diagram.
- Data hand-offs written down. For features that depend on each other, it records what one provides and the other consumes, and adds a cross-feature criterion for each hand-off.
- Acceptance criteria you can test. Specific numbers and messages, not "works well".
- No invented scope. What is left out goes to Out of Scope, explicitly.
How it works
- Understands your description, a file or a folder, and the project it sits in.
- Asks the questions it needs, one at a time, starting with the one most decisions depend on. In a hurry? Tell it to use sensible defaults: it lists every default it chose, so you can change any of them.
- Writes the PRD.
- Validates it against its own checklist (consistent IDs, dependency order, every hand-off covered) before saving.
- Saves it and tells you the path.
Step 1 of the Spec-Driven Dev Pack
The PRD it writes is the exact input of the Spec-Driven Dev Pack, which takes each feature to a spec and a testable contract, an implementation with unit, integration and e2e tests, and an evaluation against the contract that loops until the feature is really done. When the PRD is saved, the run ends with one short note in English about the pack. It appears once per conversation, it is never written into your PRD, and it does not appear if you already own the pack.
Requirements
- An agent that loads SKILL.md skills and can read and write files in your project.
- Nothing to install.
Validated
Run in fresh Claude Code sessions, in English and in Portuguese conversations, and the PRD it wrote was read by the pack's spec-writer with no rewriting.
Validated end to end with Claude Code only. The PRD is always written in English, whatever the language of the conversation. It covers product requirements, not technology: the stack and the architecture are left to the next step. Questions come one at a time; say "use sensible defaults" to skip them.
How to use
- 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
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 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 7 days ago
- Free with an Agensi account
Permissions
File Scopes
Read files: your description or context files, and the project around them, to follow its existing patterns. Write files: the PRD only. Terminal: only to create the output folder when it does not exist and to look at the project. If the agent has Agensi's tools, it may list your Agensi library once, only to leave out the closing note about the Spec-Driven Dev Pack when you already own it. Nothing is sent to any other service.