descript-api_skill

This skill automates Descript project creation, media import, and Underlord edits via API, saving time by handling end-to-end podcast workflows.
  • HTML

4

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

3 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill cdeistopened/opened-vault --skill descript-api

  • SKILL.md15.6 KB

Overview

This skill automates Descript project creation and editing via the Descript API (Early Access). It imports media from public URLs, auto-transcribes, and lets you issue one-shot natural-language edit requests to the Underlord agent. You can poll job status and retrieve results programmatically without opening the Descript app. Note: several endpoints and features are early access or WIP.

How this skill works

Use the import endpoint to create a new project and upload publicly accessible media (supports multitrack definitions). Send a single natural-language prompt to the agent endpoint to apply edits like removing filler words, adding Studio Sound, or creating highlights. Poll the job endpoint to monitor progress and inspect result metadata such as created compositions, media status, and agent_response.

When to use it

  • Automating podcast or interview imports and rough edits from hosted recordings
  • Batch-creating new Descript projects from presigned S3/GCS links or public file URLs
  • Applying consistent, repeatable edits (filler removal, audio enhancement) across episodes
  • Generating highlight reels or bulk captioning without manual app interaction
  • Integrating Descript actions into CI/CD or content pipelines

Best practices

  • Provide fully signed or publicly accessible URLs that support HTTP Range requests and last 12–48 hours
  • Frame agent instructions as one-shot prompts containing all required details—no multi-turn editing
  • Poll job status and handle partial/error outcomes; check media_status per file
  • Use short, explicit prompts for timestamps or structural cuts to reduce ambiguity
  • Expect limitations: cannot import into existing projects, some visual ops may fail, and word-level edits are limited

Example use cases

  • Ingest a raw remote recording, create a multitrack project, then remove filler words and add Studio Sound automatically
  • Generate a 30-second highlight reel from a recorded webinar using a single agent prompt
  • Batch create projects from presigned S3 links and kick off transcription for a content backlog
  • Apply captions and export project links after automated edits complete, with webhook notification

FAQ

Yes. URLs must be publicly accessible or presigned and support HTTP Range requests; sign URLs for 12–48 hours to avoid expiration during processing.

Can I edit an existing project?

Not yet. The import endpoint always creates a new project; importing into existing projects is currently unsupported.

Can the agent accept multi-step conversational edits?

No. Underlord accepts one-shot natural-language prompts only; include all required instructions in the single request.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational