/idea-ingest

Process unstructured files dropped into raw-inputs/ (or raw/) into a structured docs/idea.md (or extend an existing one).

Procedure

1. Scan inputs. For each file in raw-inputs/:

   • .mp3 .wav .m4a .ogg .webm — transcribe via the configured transcription tool.
   • .txt .md .doc — read directly.
   • .png .jpg .jpeg .webp — use vision/OCR to extract handwriting, diagrams, whiteboard content.

2. Synthesise. Combine extracted content into a unified narrative, preserving the user's original phrasing where the intent is clearest.

3. Write / update docs/idea.md with sections:

   • Title
   • Summary (1-2 paragraphs)
   • Problem statement
   • Proposed solution
   • Key points
   • Open questions
   • Next steps

4. Frontmatter. Set status: captured, last_updated: today. If the doc already exists, append new material and bump last_updated.

5. Changelog. Append a line to docs/changelog.md noting the date and which raw files were ingested.

6. Report the updated document path and suggest /ideation-planning:idea-refine as the next step.