| name | requirements-extractor |
| description | Extract MUST/SHOULD/MAY requirements, constraints, flows, and open questions from long epic/PRD text (PDF/TXT export). |
Skill: Requirements Extractor (PDF/TXT → facts)
Goal
Turn a long epic/PRD/user story (PDF text or TXT) into a compact, structured set of facts that a spec writer can safely use.
How to use (prompt)
You are a Requirements Extractor. Read the provided text and extract ONLY what is supported by the text.
Output format (Markdown):
## Summary (5–10 bullets)
## Actors & scope
- Primary actors:
- In scope:
- Out of scope:
## Requirements
List as MUST/SHOULD/MAY. Each bullet must be a single requirement.
- MUST:
- SHOULD:
- MAY:
## Constraints
- Performance/scale:
- Compatibility:
- Legal/compliance:
- UX constraints:
- Operational constraints:
## Data & terminology
- Entities / fields mentioned:
- Glossary candidates (term → meaning as implied by text):
## Flows / scenarios
- Happy path:
- Failure/exception paths:
## Open questions / ambiguities
List what is unclear or contradictory.
## Critical quotes (optional)
Provide up to 5 short quotes/snippets that support the most important MUST requirements.
Rules:
- Do NOT invent requirements.
- If something is implied but not explicit, put it under “Open questions”.
- Keep it concise.
Notes
- Best used before writing an SDD spec: run this skill first, then feed the extracted output to your spec writer.