wavespeed-dsh-skill

When a DeepSeek Harness agent needs to generate or edit images, videos, audio, or 3D assets, it often lacks a smooth calling pipeline. wavespeed-dsh-skill fills that gap by connecting the agent directly to the WaveSpeed platform, turning model discovery, schema inspection, and generation into a single natural workflow.

Under the hood the skill builds on the open-source @wavespeed/cli, where every model on the platform is one wavespeed run call. The agent follows a find → inspect → run pattern: browse the live catalog, read a target model's input schema, fill in parameters, and execute. Local files can be uploaded with the @path marker, and the skill also covers project aliases (wavespeed.json), price and balance checks, and prediction history—letting the agent go from model selection to final output in a single conversation.

If you are building creative pipelines in DeepSeek Harness, or want other agents such as Claude Code, Cursor, and Codex to tap into WaveSpeed's media generation capabilities, this skill drops in seamlessly. It ships under the MIT license, consistent with the CLI, and can be committed to a shared repository for team-wide use.