prompt-array-to-codex-images-to-grok-videoslisted

# Prompt array to Codex images to Grok videos Use M8M flow `prompt_array_to_codex_images_to_grok_videos_v2` by default. It accepts exactly one absolute `phase1_handoff_path` using `article2broll_phase1_handoff_v1`, with optional `phase1_handoff_sha256`, or one legacy `prompts` array. A handoff appends to its exact checksum-bound project; prompt-only mode creates a clearly labeled standalone project and infers no Design or Knowledge gem. Preflight Codex built-in ImageGen, Grok Build cached login and privacy/ZDR route, and bundled FFmpeg/FFprobe before generation. Never install, authenticate, change privacy, provision storage, expose credentials, use xAI REST, or switch providers. The automatic phases are `media_request_ready`, model-inspected `codex_images_generated_frozen`, `phase2_stills_published`, style-neutral `image_video_prompts_frozen`, checkpointed `grok_videos_generated_frozen`, capsule-authorized `phase3_video_evidence_frozen`, sampled-frame model-inspected `phase3_video_qa_passed`, and `media_delivery_published`. There are no user approval pauses. Use fresh no-history workers from exact isolated M8M capsules. Keep the exact approved Traditional Chinese headline; `／` is not rendered. Never add a fixed Klein Blue, paper-collage, halftone, or `no text/no letters` rule. Derive style from the frozen Design gem and still. The approved headline stays locked, unchanged, and legible; forbid only new/replacement/distorted/additional text. Retry failed still IDs only, up