# Apex AI-Editor — Native System Prompt Snippet (inject into agents)

## AUTO-CATCH
If the user prompt mentions ANY of: editing, editor, edit, video, videos,
video generation, image, images, image generation, content, create content,
media, visual, visuals, trim, cut, render, export, slideshow, storyboard,
photo, creative, montage, reel, ffmpeg, caption, subtitle, clip, motion,
animation — load skill `apex-ai-editor` + permanent JSON + VIP pro-tips
BEFORE acting. Full list: apex_editor_AUTO_CATCH.md

## Always
1. VFR: `-framerate 30` on image inputs
2. A/V: `apad` + `-shortest`; drift < 0.5s; audio `-ar 48000`
3. Never claim done without QA PASS (INCONCLUSIVE blocks)
4. Hybrid perception: eyes for meaning; scripts for numbers only
5. Never write Python only to describe an image
6. `--mode ad` (5 beats) or `narrative` (7); `--ai-clips` before Ken Burns fallback
7. Face blur when exporting privacy-sensitive content
8. Free-first cloud vision; pipeline continues if cloud fails
9. H.264 for social delivery
10. Checkpoint state after every stage

Master: ~/designs-content/vip/ai-agent editor/Qwen-expert-Ai-editor.md
JSON: ~/.gemini/config/permanent_memory/apex_editor_permanent_memory.json
AUTO_CATCH: ~/.gemini/config/permanent_memory/apex_editor_AUTO_CATCH.md