home/roleplay/visual-direction-4

Visual Direction

GPTClaudeGemini··361 copies·updated 2026-07-14
visual-direction-4.prompt
# Visual Direction Prompt Policy

Stage id: `visual_direction`
Owner: RedCube AI
Purpose: define visual language, rhythm, density, layout risk controls, visual anchors, and anti-template constraints before artifact creation.

Canonical policy:
- Build visual direction from current approved blueprint or note plan, not from fixed examples.
- Cover all real page ids, rhythm roles, peak pages, layout family ceilings, typography, spacing, and forbidden regressions.
- Preserve source-language discipline and audience-facing wording boundaries.
- Return visual direction refs and blockers; do not write final artifact blobs.

Professional skill routing:
- Route PPT visual language, rhythm, density, peak pages, layout variety, and anti-template constraints to `agent/professional_skills/rca-ppt-visual-director/SKILL.md`.
- Route template/reference deck profiling, semantic zones, placeholder capacity, and layout grammar first to `agent/professional_skills/rca-template-profiler/SKILL.md`; visual direction consumes that profile instead of re-profiling it.
- Keep `agent/skills/*.md` as stage skill policy refs only; they do not replace professional specialist skills.

Detailed prompt locators:
- `ppt_deck`: `prompts/ppt_deck/visual_direction.md`
- `xiaohongshu`: `prompts/xiaohongshu/visual_direction.md`
- `poster_onepager`: `prompts/poster_onepager/visual_direction.md`

Authority boundary:
- Visual direction acceptance is an AI-first RCA judgment surface.
- Mechanical layout metrics may inform blockers, but cannot replace director judgment.

when to use it

Community prompt sourced from the open-source GitHub repo gaofeng21cn/redcube-ai (Apache-2.0). A "Visual Direction" style prompt — adapt the placeholders and specifics to your task. Imported as-is and not independently retested here, so check the output before relying on it.

tags

roleplaycommunitygeneral

source

gaofeng21cn/redcube-ai · Apache-2.0