Guide · Agent skill
How to use the Requirements Writer agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
Requirements Writer is an agent skill that turns a product manager's rough notes about a feature into one structured SPEC.md file. It is written for product managers, and for developers or AI agents who need unambiguous requirements to build from. After this guide you will know what the skill does, how to install it, how to trigger it with a plain request, and what to check before you trust the output.
Requirements Writermvschwarz/openrig · Apache-2.0 Download the skill Quick start
- 1Download the ZIP. Get the requirements-writer skill ZIP from this directory page.
- 2Unzip into a skills folder. Place the requirements-writer folder in ~/.claude/skills/ for personal use or in your project's .claude/skills/ folder.
- 3Or upload in the Claude apps. In the Claude apps, go to Settings → Capabilities → Skills and upload the skill.
- 4Gather your inputs. Collect your rough notes and any validation.md, background.md or existing SPEC.md for the slice.
- 5Ask in plain language. Describe the feature and ask the agent to turn your intake into a SPEC.md.
- 6Answer the rounds of questions. Respond to the targeted questions; the agent will tag unconfirmed items as [draft].
- 7Review the SPEC.md. Check the GIVEN/WHEN/THEN criteria, scope lists and proof checklist before handing the file to a developer or agent.
What is an agent skill, and what does this one do?
An agent skill is a folder containing instructions (here, a SKILL.md file) that an AI agent loads when your request matches the skill's description. This skill is named requirements-writer and its description reads: use it when converting PM intake into a slice SPEC.md with observable acceptance outcomes, explicit scope, and advisory sibling build dependencies.
Once active, the agent acts as a product analyst. It interviews you through a focused, multi-round conversation about scope, users, behavior, business rules and edge cases. It then produces a single SPEC.md with GIVEN/WHEN/THEN acceptance criteria, explicit in-scope and out-of-scope lists, and a proof checklist. Items you have not confirmed are marked [draft].
- Owns the what and why of a feature
- Does not make architecture, implementation, timeline, estimation, data model or API decisions
- Checks for existing validation.md, background.md and requirements files before asking questions
- Keeps the output to current-phase scope; future phases go under Out of Scope
How do you install the Requirements Writer skill?
Download the ZIP of the requirements-writer skill from this directory page. The package contains SKILL.md. Unzip it so the requirements-writer folder sits in one of two places:
You can install it for yourself across all projects by placing it in ~/.claude/skills/, or for a single project by placing it in that project's .claude/skills/ folder.
In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: ~/.claude/skills/requirements-writer/
- Project install: .claude/skills/requirements-writer/ inside your project
- Claude apps: Settings → Capabilities → Skills, then upload the skill
How do you trigger the skill with a natural request?
You do not need a special command. The skill loads when your request matches its description, so describe the job in plain words that mention converting intake or writing a spec. Give the agent your rough notes in the same message or the next one.
Useful triggers include asking the agent to turn your intake into a SPEC.md, to write acceptance criteria for a feature, or to draft requirements for a slice. Mention the feature and paste whatever you have, even if it is messy.
- Turn this PM intake into a SPEC.md for the notifications slice.
- Write GIVEN/WHEN/THEN acceptance criteria for a password reset feature from these notes.
- Interview me about this feature and produce the slice spec with in-scope and out-of-scope lists.
What happens during the conversation?
The first round starts with the agent summarizing what it understands in two or three sentences, mapping the feature to your existing product, and asking five to eight focused questions. Later rounds cover scope and personas first, then acceptance criteria, business rules and edge cases, and finally scope refinements and open questions.
After each exchange the agent returns the current state of the requirements. Anything still needing your input is tagged [draft]. If you are unsure about an answer, the skill is designed to offer two or three concrete options with trade-offs.
What does the SPEC.md output look like?
The output is a markdown file with front matter (id, title, status, owner, intent, depends_on) followed by Intent, Mini-requirements (personas, user stories, acceptance criteria grouped by functional area, business rules), Scope (in and out), Open Questions and a Proof contract checklist.
Acceptance criteria use GIVEN for the starting state, WHEN for the trigger and THEN for the observable result. Each criterion should stand alone and describe what the user sees rather than internal system behaviour. The depends_on field lists only same-parent dot-IDs and is advisory build order; a stale or missing edge is reported by graph readers but never blocks work.
- Front matter: id, title, status, owner, intent, depends_on
- Sections: Intent, Mini-requirements, Scope, Open Questions, Proof contract
GIVEN [context or precondition]
WHEN [user action or system event]
THEN [expected observable result] — [draft] if not yet confirmedWhat do you need before you start, and what are the safety notes?
The skill lists no requirements. It does not run scripts and does not use the network. It consists of a single SKILL.md file, so there is nothing to execute on install.
The skill works best if you have validation.md (office hours output) or background.md, and any existing SPEC.md or legacy requirements.md for the slice. If validation.md has a GO verdict, the agent skips demand and scope questions and moves straight to acceptance criteria. If only a legacy requirements.md exists, the skill preserves it as input and reconciles it into SPEC.md.
- Requires: nothing listed
- Runs scripts: no
- Uses network: no
- Helpful inputs: validation.md, background.md, existing SPEC.md or requirements.md
What people use it for
- Turning a product manager's messy notes from a feature intake into a single SPEC.md for a slice
- Drafting GIVEN/WHEN/THEN acceptance criteria for a feature so a developer or AI agent can build from them
- Separating in-scope from out-of-scope work for the current phase, with future phases listed under Out of Scope
- Producing a proof checklist of observable outcomes to confirm a slice worked
- Reconciling a legacy requirements.md into the SPEC.md format
Frequently asked questions
Is the Requirements Writer skill free?
The source material for this page lists the licence as Apache-2.0. It does not state prices, so check the repository for any cost details before you rely on it.
Does the skill need any setup or dependencies?
It lists no requirements, runs no scripts and uses no network. Installation consists of placing the requirements-writer folder containing SKILL.md in the correct skills directory.
Where do I put the requirements-writer skill folder?
Put it in ~/.claude/skills/ to use it across projects, or in a project's .claude/skills/ folder to limit it to that project. In the Claude apps you can upload it under Settings → Capabilities → Skills.
Why does the skill refuse to suggest architecture or estimates?
Its boundaries say it owns the what and why only. It does not make architecture or implementation decisions, estimate timelines or effort, define data models or API contracts.
What if my requirements are only rough notes?
That is the intended input. The agent summarizes what it understands, asks five to eight focused questions in the first round, and marks unconfirmed items as [draft].
Does it work if my team does not use SPEC.md files?
The output template is built around the SPEC.md and slice conventions. Projects not using them would need to adapt the template.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
Use Requirements Writer if you are a product manager who wants a clean, implementable SPEC.md from rough notes, or a team that keeps one requirements file per feature in slice-style workflows. Skip it if you only need a quick brainstorm, if your project does not use SPEC.md conventions, or if you want help with technical design or estimation, which the skill explicitly refuses to do.
Download the skill