pks prd

pks prd is the Product Requirements Document branch of the pks CLI: six subcommands that write a conventional PRD skeleton to docs/PRD.md, parse it back, and check that its sections are filled in. It is a document scaffolder and a structural linter — the writing itself stays yours.

Overview

The branch registers under pks prd and shares one settings class across every subcommand, so --verbose, --output-format, --config, and --no-color parse on all six. Two subcommands write files (template, generate), three read one (load, status, validate), and one lists requirements (requirements).

• Write a skeleton. template emits a static Markdown archetype with bracketed placeholder prompts. generate does the same and seeds a few sections from words it finds in your idea sentence.
• Read it back. load parses a Markdown or JSON PRD and prints project name, template, and section count.
• Check it. validate reports missing pieces and exits non-zero when the document fails, which makes it usable as a pre-commit or CI gate.

Note. The branch description in the CLI says "AI-powered generation". No subcommand calls a language model. generate matches keywords in your idea string; template substitutes three placeholders into a fixed string. Treat this branch as scaffolding and structural checking, not authoring.

What you get

• Seven template archetypes. standard, technical, mobile, web, api, minimal, and enterprise, accepted case-insensitively by both template --type and generate --template.
• A conventional section layout. Overview, Business Context, Requirements, User Stories, and the rest of the shape reviewers expect, so a new repository starts from a recognizable document.
• A CI-usable gate. pks prd validate exits 1 when the document has errors, and --strict promotes a missing description and an empty requirements list from warnings to errors.
• A JSON side door. load --export, status --export, validate --report, and requirements --export each write machine-readable output to a path you choose.
• A directory sweep. status --check-all walks the current directory for PRD*.md, prd*.md, *requirements*.md, *spec*.md, and *.prd.json and tables what it finds.

How it fits together

Everything centers on one file path. template and generate both default their output to ./docs/PRD.md, creating docs/ when it is missing. load, status, validate, and requirements all default their input to the same path, so a bare pks prd status in a repository root works with no arguments once a PRD exists.

The parser recognizes two shapes: a JSON PrdDocument, or Markdown whose first heading matches # <Name> - Product Requirements Document and whose sections are ## headings. The Markdown path extracts titles and sections but never populates a requirements array, which is why the numbers you see downstream are structural rather than semantic.

• Trustworthy: section counts, file presence, template type, project name, validation errors and warnings.
• Not trustworthy: completion percentages from status, the requirement rows from requirements, and the clarity, consistency, and feasibility scores from validate.

Commands

generate · template · load · validate · status · requirements

Subcommand	What it does	Trust level
template	Writes a static PRD skeleton for one of seven archetypes.	Deterministic.
generate	Writes a PRD skeleton seeded from an idea sentence.	Keyword-driven, not AI.
load	Parses a PRD file and prints a summary panel.	Structural facts only.
validate	Reports completeness errors and exits non-zero on failure.	Structural checks only.
status	Prints counts and a breakdown chart, or sweeps the tree.	Counts real, completion synthetic.
requirements	Lists and filters requirements with CSV or JSON export.	Placeholder rows — see its page.

Next steps

• Generate a PRD from an idea — the fastest way to get a filled-in starting document
• PRD templates — the seven archetypes and how to pick one
• Validate a PRD — wire the completeness gate into CI
• PRD status and progress — counts, the directory sweep, and what the percentage really means
• Load and parse a PRD — the parser's exact expectations for headings
• pks prd CLI reference — every flag on all six subcommands in one table