← ClaudeAtlas

storyboardlisted

Plans one episode and stops for approval before generation. Use when the user asks to "스토리보드 만들어", "스토리보드 작성", "이 주제로 영상 기획", "촬영 대본 만들어", "내가 녹화할 대본", "make a storyboard", "plan a video for topic X", or starts a new topic in a channel. Researches the topic, finds the wow points, decides three messages for today's viewer, writes one seven-item scenario per message, shows all three for the pick, researches the winner, and writes the storyboard under data/[channel]/episodes/[topic]/storyboard/. Format: 9:16 shorts by default, or 16:9 long-form with chapters. The narration is written before any shot and read on its own twice — the chain, then the words — each looped to 95 in at most three reads with the sentences handed to the reviewer inline; the board gets the contract checkers and the author's own read. Plans each promise, visual change and sound event. Nothing is generated here. Produce builds the approved board; autoproduce runs unattended.
ttalkkaklab/social-flow · ★ 0 · AI & Automation · score 73
Install: claude install-skill ttalkkaklab/social-flow
# Storyboard authoring — data/[channel]/episodes/[topic]/storyboard/ A linked copy without a valid `headRevisionNo` cannot save or checkpoint: pull `mode: "side"`, merge, and pass the returned head as `baseRevisionNo`. Keep `.portal.json`; directory-free checkpoints also require an explicit base. For pending portal render requests, follow [the six-ratio allocation contract](references/production-mode.md): `portal_render_allocation` read → host LLM reasons over every shot → submit requestId/baseRevisionNo and all assignments → pull. New choices persist `PRODUCTION.renderRatioVersion:1`; neither a pending request nor ratio selection approves spending. Read [story-quality.md](references/story-quality.md) before candidates or narration. Its evidence → meaning → ending → optional CTA contract overrides older mandatory-question and modern-case rules. Write `window.STORY` in §4a; draft checks require it. The existing narration review supplies its four evidence-backed findings; after vocabulary edits revalidate the read and run `check-story.js storyboard/` before §4b or approval. No score waives a failed criterion. Takes one topic through **research → wow points → three messages → three scenario candidates → one pick → more research → narration (the story pass) → narration read-through (looped to 95) → narration vocabulary (looped to 95) → the narration approval → the board (camera · space · sound · slides) → the image and clip plan → storyboard approval**. **No generation call happe