prd-parser

Splits large PRD documents into structured FR/NFR chunk files with change tracking.

Updated Jan 12, 2026
One-click install
npx skills add https://github.com/ab300819/keel-workflow --skill prd-parser-ab300819
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: prd-parser
Source: https://github.com/ab300819/keel-workflow/tree/main/skills/prd-parser
Command: npx skills add https://github.com/ab300819/keel-workflow --skill prd-parser-ab300819

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Large product requirement documents (Markdown, PDF, or screenshots) are hard to process incrementally. This Skill parses them into structured, individually trackable chunk files while preserving the original text verbatim, enabling block-by-block requirement clarification later. ## Core Features & Use Cases - Multi-format parsing: Reads Markdown, PDF (paged, up to 20 pages per read), and image/screenshot input, then archives source files into docs/prd/source/ with a manifest. - Structured chunking with classification: Splits documents by TOC or logical topics into 30-300 line chunks, pre-classifying each as FR (functional) or NFR (non-functional) with continued numbering. - Change detection via dual fingerprints: Uses document-level and chunk-level sha256 fingerprints plus stable chunk_key anchors to detect updates, marking chunks as outdated, pending, or removed without renumbering. - Use Case: You receive a 50-page PRD PDF. Run this Skill to split it into FR-01 through FR-08 chunk files under docs/prd/chunks/, then hand them to prd-brainstorm for per-block clarification. ## Quick Start Ask the AI to parse your PRD file, for example: run /prd-parser on docs/my-product-prd.pdf to split it into classified requirement chunks.

Frequently Asked Questions about prd-parser

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I split a large PRD document into structured requirement chunks?

Run the prd-parser skill on your PRD file. It detects the document structure via table of contents or logical topics, splits content into 30-300 line chunks, and writes each as a Markdown file with YAML frontmatter under docs/prd/chunks/.

What file formats does PRD parsing support?

Markdown, PDF, and screenshots/images are natively supported. PDFs are read in pages of up to 20 at a time, and images are processed via OCR. Word docx files are not supported and must be converted to Markdown or plain text first.

How does PRD change detection work when the document is updated?

The parser computes a document-level sha256 fingerprint to detect any change, then matches chunks by stable chunk_key anchors and compares per-chunk fingerprints. Changed chunks are marked outdated, new sections become pending, and deleted sections are marked removed without renumbering.

Does the parser rewrite or interpret requirement content?

No. The parser is explicitly constrained to preserve original text verbatim with no rewriting, deletion, or interpretation. Requirement clarification and FR/NFR reclassification are delegated to the separate prd-brainstorm skill.

How are FR and NFR requirements classified and numbered?

Each chunk is pre-classified as FR (functional: user operations, business logic) or NFR (non-functional: performance, security, compliance). Numbering continues from the highest existing ID in docs/prd/, starting at FR-01 or NFR-01 when the directory is empty.