What problem does it solve? Getting raw multimodal content — voice memos, screenshots, PDFs, YouTube videos — into a personal knowledge vault requires manual transcription, OCR, and note formatting. This Skill automates that intake pipeline so every capture lands in the vault inbox as a properly frontmattered fleeting note. ## Core Features & Use Cases - Multimodal capture: Detects input modality (voice, image, PDF, YouTube) from file extension or URL and selects the right transcription or OCR tool (whisper-cpp, tesseract, pdftotext, yt-dlp) with fallback support. - Approval-gated writes: Presents the verbatim transcript for user approval before anything is written to the vault, with a feedback loop to retry with a different tool. - Provenance-rich notes: Writes fleeting notes to the vault inbox with status: draft, type: fleeting, source_type, and captured_at so downstream skills like process-notes can atomize them later. - Use Case: Record a voice memo on your phone, drop the .m4a file into your project, and have it transcribed, reviewed, and filed as an inbox note ready for later atomization into permanent notes. ## Quick Start Capture the attached voice memo into my Obsidian vault as a fleeting note using vault-ingest.