OpenParable
Skip to content
← Back to skills

Video Timeline

Documents
video-timeline

Video Timeline is an Agent Skill for documents published in the majiayu000/claude-skill-registry repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Generate and validate timeline.json using a builder + validator team.

View on GitHub ↗
QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

  1. Open your project in Codex.
  2. Copy and paste the prompt below.
  3. Review files and risks before approving any action.
Prompt to paste
Inspect the Agent Skill at https://github.com/majiayu000/claude-skill-registry/blob/HEAD/skills/orchestration/video-timeline/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/video-timeline, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Video Timeline Pipeline

Generate and validate timeline.json using a builder + validator team. The builder generates the timeline, the validator enforces quality gates. Max 3 validation rounds.

Team Structure

Create an agent team called "timeline-edit" with these teammates:

builder (general-purpose agent)

Role: Generate timeline.json from script + transcript + assets. Skills to preload: Load the timeline-builder agent instructions. MCP servers needed: giphy, pexels Tasks:

  1. Read the prompter/script file (or READ each slide image if no prompter exists)
  2. Read the transcript (transcript-clean.json > transcript.json > SRT)
  3. List available slides, GIFs, and B-roll in the video directory
  4. If GIFs are missing or below density minimum, search and download via Giphy MCP
  5. If B-roll is missing or below density minimum, search and download via Pexels MCP
  6. Match slides to speech moments using content understanding
  7. Generate creative edit decisions with ~20% distribution per layout type
  8. Write timeline.json to the video directory
  9. Mark task complete and notify the lead

validator (general-purpose agent)

Role: Validate timeline.json against 21 quality checks. Skills to preload: Load the timeline-validator agent instructions. Model preference: haiku (cheap and fast - just counting and checking rules) Tasks (run AFTER builder completes):

  1. Read timeline.json from the video directory
  2. Read transcript-clean.json (or transcript.json)
  3. List files in slides/, gifs/, broll/ directories
  4. Run all 21 quality checks (structural, density, speech alignment)
  5. Output PASS or FAIL verdict with specific fix instructions

Workflow

  1. Create team "timeline-edit"
  2. Create task "Build timeline" and assign to builder
  3. Builder generates timeline.json, marks task complete
  4. Create task "Validate timeline" and assign to validator
  5. Validator reads timeline.json + transcript, runs 21 checks
  6. If PASS: clean up team, report success
  7. If FAIL: create task "Fix timeline" for builder with the validator's specific feedback
  8. Builder applies targeted fixes, marks task complete
  9. Create task "Re-validate timeline" for validator
  10. Max 3 rounds (build + validate). If still failing after 3, report remaining issues to user.

Input

Provide the video project directory:

$ARGUMENTS