AgentSkillsCN

step5-shots-prompts

在 AIMVDashboard 中填写第 5 步(镜头/提示),通过整理镜头提示、为其建立索引,并在“镜头”页面中验证其就绪状态。在被要求完成或核对第 5 步提示相关成果时使用。

SKILL.md
--- frontmatter
name: step5-shots-prompts
description: Fill out Step 5 (Shots/Prompts) for AIMVDashboard by compiling shot prompts, indexing them, and validating readiness in the shots page. Use when asked to complete or verify Step 5 prompt artifacts.

Step 5 Shots & Prompts Skill

Complete prompt outputs so the Step 5 page has fully usable shot content.

Required artifacts:

  • Shot prompt files under projects/<project-id>/prompts/ (Kling/Nano Banana/Suno as applicable)
  • projects/<project-id>/prompts_index.json

Execute

  1. Resolve project id (default default).
  2. Ensure Step 3 canon data is complete before prompt compilation.
  3. Load and apply references/system-prompt-best-practices.md for all systems used in this project (Kling, Nano Banana, Suno, SeedDream).
  4. Generate or update shot prompts for all required shots/options using system-specific best-practice structure.
  5. Regenerate prompt index:
bash
npm run index -- --project <project-id>
  1. Ensure each shot expected by shot_list.json appears in prompts_index.json.

System-specific quality checks

Before final save:

  • Kling prompts include camera/framing/movement + negative constraints.
  • Nano Banana prompts are static-frame descriptions (no motion timeline language).
  • Suno prompts describe music only (no visual/camera terms).
  • SeedDream prompts include composition + rendering intent + exclusions.

Completeness check

bash
project=default
node -e 'const fs=require("fs");const idx=JSON.parse(fs.readFileSync(`projects/${process.env.P}/prompts_index.json`,`utf8`));if(!Array.isArray(idx.shots)||idx.shots.length===0){console.error("No shots in prompts index");process.exit(1)}console.log(`shots indexed: ${idx.shots.length}`)' \
  P="$project"

Step 5 is complete only when the index is populated and representative prompt files exist for indexed shots.