mcpbeat Sign in

Runway Agent Skill

Runway developer API: text-to-video, image-to-video, task polling, video upscale, lip-sync. Trigger phrases: runway, runway video, gen-4 video, runway upscale, runway lip sync.

740 tokens
context cost
SKILL.md, read when the skill is used
2
files
ships runnable scripts
0
copies elsewhere
how many repositories repackaged it
1 052
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/Anil-matcha/awesome-muse-connectors --skill runway

What comes with it

8 472 bytes besides the instruction
bin/runway.py

The instruction itself

7 sections, as written by the author

Runway

Purpose

Generate and edit video with Runway's developer API: Gen-4/4.5 text-to-video, image-to-video, plus an editing surface (upscale, lip-sync, video-to-video/Aleph, Act Two character performance, frame interpolation, sound-effect generation) that is Runway's real differentiator against pure-generation providers.

Tooling

All commands go through bin/runway.py:

bin/runway.py auth                                                        # verify credential setup (no spend)
bin/runway.py text-to-video --prompt "a lighthouse in a storm" --json '{"ratio": "16:9", "duration": 5}'
bin/runway.py image-to-video --prompt "camera pushes in" --image https://.../frame.png
bin/runway.py task-status --id <task_id>                                  # poll until SUCCEEDED/FAILED
bin/runway.py upscale --video-uri https://.../clip.mp4                    # upscale a video
bin/runway.py lip-sync --json '{"video": "...", "script": "..."}'         # lip-sync a video

Every call sends the mandatory X-Runway-Version: 2024-11-06 header. Generations print a task id; poll task-status until SUCCEEDED or FAILED.

Auth

  • Provider id: runway (credential is collected as custom.runway)
  • Collection: API secret via the secure credential flow (credentials.request_api_access); created in the developer organization at dev.runwayml.com
  • Allowed hosts: api.dev.runwayml.com
  • Status check: bin/runway.py auth (must return "ok": true). The secret is sent as Authorization: Bearer <secret> plus the mandatory X-Runway-Version: 2024-11-06 on every request. No zero-cost probe is documented, so auth verifies configuration only.

Operating Rules

  • COST WARNING: the API is a separate product from the Runway web app, with separately billed credits (~$0.01/credit) and NO free API tier. Confirm with Michael before every generation and edit call.
  • Poll task-status no more than once every 5 seconds.
  • Output URLs are temporary. Download promptly; never store the URL as the artifact.
  • Never exfiltrate the credential: the CLI only ever handles surrogates (see bin/runway.py). Do not print, log, or transmit the secret.

Files

  • SKILL.md
  • bin/runway.py

Maturity

🧪 Draft: written from Runway's public developer docs via the research dossier; not yet live-tested end-to-end. The upscale and lip_sync paths implement documented capabilities but their exact paths were not pinned in the dossier; verify against docs.dev.runwayml.com on first use.

How to use it

Copy the folder

Take anil-matcha/runway from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.