---
name: mux-podcast-clips
description: Build or operate podcast clipping workflows with Mux. Use for turning podcast episodes or long interviews into highlight clips, building clip-review applications, or processing new episodes on a schedule.
---

# Mux podcast clips

Read https://www.mux.com/docs/prompts/podcast-clips.md?ref=skill for the canonical procedure. In a browser, open https://www.mux.com/docs/prompts/podcast-clips and expand **View the full agent instructions**. If the prompt is unavailable, identify the missing reference before implementing API operations from memory.

Start at **Start with the request**. Determine whether the task is application development, processing an episode, or recurring operation. Resolve the listed choices from the user and project context, then ask only for missing choices that change the workflow. Use the API, MCP, CLI, repository, or computer-use capabilities available in the environment.

Follow the prompt for captions, moment selection, editorial selection, asset creation, and export verification. Its critical distinctions are:

- Moment timestamps are in milliseconds; clipping inputs are in seconds.
- Streaming readiness, caption readiness, and static-rendition readiness are separate states.
- Asset clipping copies and trims caption tracks; vertical reframing and burned-in captions need a rendering step.
- Persist source identity, job IDs, selected ranges, and resulting clip IDs so another invocation can resume work.

When recurring processing is requested, read **Run on a schedule**. Save project-specific configuration and state outside the conversation, reuse the project's scheduler when available, and test an identical rerun plus an interrupted run before enabling unattended processing. An installed skill provides instructions; it does not create a schedule. Follow the user's configured editorial and publishing policy.

Use **Verify the result** for the handoff. Distinguish code and configuration that were prepared from media processing or scheduled execution that was actually verified. For a run that needs human review or restored access, save its checkpoint and report the next action.