prd

Run the prd skill with your idea as input. It detects short unstructured input and routes to brainstorm mode, which uses 5W1H, user journey mapping, and MoSCoW prioritization to produce structured FR-XX/NFR-XX requirement files with acceptance intent.

Provide the PRD file to the prd skill, which auto-detects documents over 2000 characters and routes to PRD parsing. The parser splits the document into FR/NFR chunks, each chunk is clarified via brainstorming, and results are synthesized into a requirements index.

Brainstorm mode handles short unstructured ideas through interactive clarification. PRD parse mode handles large structured documents by splitting them into chunks first, then clarifying each chunk. Inputs between 200 and 2000 characters trigger a confirmation question.

Yes, use the --revise flag with a specific FR number, such as --revise FR-03. The skill re-enters brainstorming for that FR only, keeps its ID unchanged, and marks any existing downstream mapping as outdated so it can be re-imported.

The docs/prd directory is a single-requirement scaffold cleaned up after close. The clear command runs a dry-run impact analysis first, flags orphaned or unsynced requirements, and only deletes after explicit confirmation.

Use requirements coding directly when your needs are already clearly defined. The prd skill is for the earlier stage: exploring fuzzy ideas or decomposing large PRD documents into structured FR/NFR items that requirements coding can then consume.