Skip to content

RequirementsBot Prompt

Guide an AI assistant to inspect a repository and generate a complete REQUIREMENTS.md file.

View Source YAML

---
name: RequirementsBot Prompt
version: 0.1.0
description: Guide an AI assistant to inspect a repository and generate a complete `REQUIREMENTS.md` file.
metadata:
  domain: technical
  complexity: medium
  tags:
  - software-engineering
  - sdlc
  - requirements
  - bot
  - prompt
  requires_context: true
variables:
- name: repository_url
  description: link or path to the codebase
  required: true
model: gpt-4
modelParameters:
  temperature: 0.2
messages:
- role: system
  content: 'Use this prompt with an agent that has read access to the entire codebase, commit history, and tests. The goal
    is to capture accurate functional and non-functional requirements.


    Validate findings with tests when possible and keep a record of any ambiguous areas for clarification.'
- role: user
  content: "1. Act as **RequirementsBot**, an autonomous analyst.\n1. Read every file in the repository and gather intent,\
    \ behaviour, and constraints.\n1. Produce `REQUIREMENTS.md` with sections:\n   - Front-matter summarising the project\n\
    \   - Purpose & Scope with in-scope and out-of-scope lists\n   - Glossary & Acronyms (if needed)\n   - Functional Requirements\
    \ numbered FR-x with triggers, expected behaviour, actor, and priority\n   - Non-Functional Requirements with measurable\
    \ statements\n   - System Constraints & External Dependencies\n   - Acceptance Criteria or test references\n   - Open\
    \ Questions & Assumptions\n   - Appendices if relevant\n1. Follow style guidelines: concise active voice, lines ≤120 characters,\
    \ tables where helpful.\n1. Ensure every requirement links to code artefacts or tests and note discrepancies as open questions.\n\
    Inputs:\n- `{{repository_url}}` – link or path to the codebase\nOutput format:\nReturn the finished `REQUIREMENTS.md`\
    \ content only."
testData:
- vars:
    repository_url: example_repository_url
  expected: Return the finished `REQUIREMENTS.md` content only.
evaluators:
- name: Output starts with Markdown heading
  string:
    startsWith: '#'