Skip to main content
skill CX Codex

requirements-gathering — Claude Code skill: collect requirements

Walks your agent through collecting and documenting stakeholder requirements before development starts. Useful at project kickoff and feature planning.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add aj-geddes/useful-ai-prompts --skill requirements-gathering -a claude-code
Open on GitHub

What requirements-gathering does

The skill gives the agent a sequence for requirements work: identify who affects the project, pull needs out of them, write everything down, and get sign-off. Five reference guides live in the references/ directory, covering stakeholder discovery, elicitation techniques, documentation, validation and sign-off, and a traceability matrix.

SKILL.md includes a Python example: a StakeholderDiscovery class with stakeholder categories and methods to find them. Full implementations sit in the reference guides, not in the skill file itself.

A separate practices section lists what to do and what to avoid: write measurable statements, define acceptance criteria, prioritise with MoSCoW, separate must-haves from nice-to-haves. Avoid vague words like quickly or easily, and avoid scope changes without a process. If your team already has its own spec format, the skill will push a different structure, so read the reference guides first.

What it helps with

  • collect requirements from every stakeholder group at kickoff
  • put verbal agreements into writing
  • build a requirements traceability matrix
  • prioritise features with MoSCoW
  • write acceptance criteria for user stories

Example prompts for the agent

  • Help me gather requirements for a new project: identify stakeholders and their needs
  • Build a requirements traceability matrix from this feature list
  • Check my requirements for vague wording and add acceptance criteria

How to install requirements-gathering

1Skills CLI (Vercel)

npx skills add aj-geddes/useful-ai-prompts --skill requirements-gathering -a claude-code

2Manually: put the files in place

~/.claude/skills/requirements-gathering/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/requirements-gathering/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about requirements-gathering

How do I install requirements-gathering in Claude?

Run npx skills add aj-geddes/useful-ai-prompts --skill requirements-gathering -a claude-code. The skill file lands at ~/.claude/skills/requirements-gathering/SKILL.md.

What does requirements-gathering do?

It guides the agent through requirements steps: stakeholder discovery, elicitation, documentation, validation and sign-off. The point is agreeing on what gets built before development begins.

Which platforms are supported?

Claude, Codex, Cursor, GitHub Copilot, Gemini CLI and OpenCode. Each platform has its own install command and skill file path.

Source text

SKILL.md · MIT
---
name: requirements-gathering
description: >
  Systematically collect, document, and validate requirements from stakeholders.
  Ensure clarity, completeness, and agreement before development begins to
  reduce scope creep and rework.
---

# Requirements Gathering

## Table of Contents

- [Overview](#overview)
- [When to Use](#when-to-use)
- [Quick Start](#quick-start)
- [Reference Guides](#reference-guides)
- [Best Practices](#best-practices)

## Overview

Effective requirements gathering establishes a shared understanding of what will be built, preventing misalignment and expensive changes later in the project.

## When to Use

- Project kickoff and planning
- Feature development initiation
- Product roadmap planning
- System modernization projects
- Customer discovery
- Stakeholder alignment sessions
- Writing user stories and acceptance criteria

## Quick Start

Minimal working example:

```python
# Identify and analyze stakeholders

class StakeholderDiscovery:
    STAKEHOLDER_CATEGORIES = [
        'End Users',
        'Business Owners',
        'Technical Leads',
        'Operations/Support',
        'Customers',
        'Regulatory Bodies',
        'Integration Partners'
    ]

    def identify_stakeholders(self, project):
        """Map all stakeholder groups"""
        return {
            'primary': self.get_primary_stakeholders(project),
            'secondary': self.get_secondary_stakeholders(project),
            'tertiary': self.get_tertiary_stakeholders(project),
            'total_to_engage': self.calculate_engagement_strategy(project)
        }

    def analyze_stakeholder_needs(self, stakeholder):
        """Understand what each stakeholder needs"""
        return {
// ... (see reference guides for full implementation)
```

## Reference Guides

Detailed implementations in the `references/` directory:

| Guide | Contents |
|---|---|
| [Stakeholder Discovery](references/stakeholder-discovery.md) | Stakeholder Discovery |
| [Requirements Elicitation Techniques](references/requirements-elicitation-techniques.md) | Requirements Elicitation Techniques |
| [Requirements Documentation](references/requirements-documentation.md) | Requirements Documentation |
| [Requirement Validation & Sign-Off](references/requirement-validation-sign-off.md) | Requirement Validation & Sign-Off |
| [Requirements Traceability Matrix](references/requirements-traceability-matrix.md) | Requirements Traceability Matrix |

## Best Practices

### ✅ DO

- Engage all key stakeholders early
- Document requirements in writing
- Use specific, measurable language
- Define acceptance criteria
- Prioritize using MoSCoW method
- Get stakeholder sign-off
- Create traceability matrix
- Review requirements regularly
- Distinguish must-haves from nice-to-haves
- Document assumptions and constraints

### ❌ DON'T

- Rely on memory or verbal agreements
- Create requirements without stakeholder input
- Use ambiguous language (quickly, easily, etc.)
- Skip non-functional requirements
- Ignore constraints and dependencies
- Over-document trivial details
- Rush through requirements phase
- Build without stakeholder agreement
- Make scope changes without process
- Forget about edge cases and error conditions

© aj-geddes · MIT · aj-geddes/useful-ai-prompts

Similar skills

All skills for Agent setup: memory, context, orchestration →

From the same set118

websocket-implementation A coding skill that walks you through building a WebSocket server: connection management, message routing, and Redis-based scaling. CX Codex +1 663installs project-estimation A skill that helps the agent estimate project scope, timeline and resources. Useful when you need a budget or a plan you can defend. CX Codex +1 659installs log-analysis A skill that helps the agent read application and system logs: spot errors, recurring patterns and the root cause of a failure. Useful when a service breaks and you need answers. CX Codex +1 657installs image-optimization A skill that walks your agent through compressing images and picking modern formats. Useful before a deploy, when pages drag on mobile networks. CX Codex +1 656installs ruby-rails-application Guides building Ruby on Rails applications with Active Record models, RESTful controllers, and JWT authentication. Best for structuring Rails APIs and managing database schemas. CX Codex +1 655installs api-rate-limiting Guides the implementation of token bucket and sliding window rate limiting. Protects APIs from brute force attacks and traffic surges using Redis or Flask. CX Codex +1 640installs information-architecture Structures content, sitemaps, and taxonomies for web apps and documentation hubs. It guides your AI agent through a four-stage process from card sorting to tree testing. CX Codex +1 634installs internationalization-i18n A skill that walks your agent through shipping a multi-language app: string extraction, translation catalogs, plurals, locale-aware dates and numbers, RTL. CX Codex +1 633installs

Project sponsors

Companies that keep this analytics open
Gold sponsor

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor