story-execute

name	story-execute
description	Execute the 9-wave story pipeline automatically. W1(story design) → W2(synopsis) → W3(writing+review) → 🛑user confirm → W4(production extract) → W5(TTS/SFX) → W6(storyboard CSV, batch QA × 3) → W7(image production, batch QA × 2) → 🛑user confirm → W8(assembly: SFX scene-match + audio import + CapCut + optional video) → W9(upload info). Each wave runs as a subagent with fresh context. Review loops max 5 rounds. Genres: yadam / dark-history / bespoke. Trigger: 'execute story', 'run pipeline', '파이프라인 실행', '자동 실행'
argument-hint	[--from W{N}] [--to W{N}]

Execute the story pipeline from W1 to W9 using wave-based subagent execution.

Each wave:

• Runs as a subagent (fresh context)
• Has built-in review loop (max 5 rounds, exits on 0 issues)
• Writes W{N}_SUMMARY.md on completion
• Updates STATE.md + W_progress.json

Manual gates (2 total):

• W3 completion → user confirms script before W4 (production extract)
• W7 completion → user confirms image QA before W8 (assembly)

Flags:

• --from W{N} — start from wave N (default: next incomplete wave from STATE.md)
• --to W{N} — stop after wave N (default: W9)

<execution_context> @skills/story-engine/workflows/execute-pipeline.md @skills/story-engine/SKILL.md </execution_context>

$ARGUMENTS

Wave reference docs are loaded per-wave by the orchestrator. STATE.md tracks current position for resume capability.

Execute the pipeline workflow from @skills/story-engine/workflows/execute-pipeline.md end-to-end. Preserve all workflow gates (review loops, user confirmation at W3 and W7, state updates).