image-inspect-for-agent

name	image-inspect-for-agent
description	Bounded EXIF/metadata and optional Tesseract OCR for one media file via scripts/image-inspect.ps1 and scripts/image_inspect.py (same stack as connectors/rich_media_sample).

Image / media inspect for agent context

When to use

• Operator points to a photo (HEIC/JPEG), audio, or video and the assistant needs tags or OCR text in the chat.
• Prefer this over inventing Pillow/Tesseract one-offs.

Ritual

1. Read docs/ops/IMAGE_EXIF_OCR_FOR_AGENT.md for L0-L3 and ML limits.
2. From repo root: .\scripts\image-inspect.ps1 -Path <file>; add -Ocr for images with visible text.
3. Keep paths under docs/private/ when handling iPhone / personal media.
4. If output is empty: check uv sync --extra richmedia, tesseract / ffprobe on PATH — report the matrix from the doc, not vague “cannot”.

Pairing

• Video screen recordings: video-frame-samples.ps1 + VIDEO_FRAME_EXTRACT_FOR_AGENT.md.
• iCloud staging: icloud-photos-fetch-range.ps1 then inspect.

Anti-patterns

• Claiming CLIP-style visual search using the product ML stack — it is text-sample detection, not photo embedding search.
• Pasting GPS EXIF or personal paths into public GitHub text.