Wavespeed

WaveSpeedAI — generic skill (any agent that follows the agentskills.io spec)

This file is intentionally portable. Copy it into whichever skill directory your host agent reads from.

Setup

export WAVESPEED_API_KEY="ws_..."     # https://wavespeed.ai/accesskey

CLI shim: ~/.local/bin/wavespeed-cli (run bash install.sh from the repo).

Trigger conditions

• User asks for AI media generation (image / video / audio / avatar).
• User mentions a wavespeed-ai/* uuid.
• User wants to upload a local file as input for image-to-X.
• User wants to call Claude / GPT / Gemini / Llama via the WaveSpeed LLM proxy.
• User asks for the WaveSpeed model catalog or balance.

Surface

wavespeed-cli run      ''    # submit + poll until terminal
wavespeed-cli submit   ''    # submit only; print task id
wavespeed-cli result                  # fetch result
wavespeed-cli cancel                  # cancel/delete
wavespeed-cli upload                     # upload local asset
wavespeed-cli models  [--names-only|--filter]  # catalog
wavespeed-cli balance                          # account balance
wavespeed-cli llm      ''       # OpenAI-compatible chat
wavespeed-cli verify-webhook        # HMAC-SHA256 webhook verify

Examples

# Text → image
wavespeed-cli run wavespeed-ai/z-image/turbo '{"prompt":"Cat in tuxedo"}'

# Image → video
URL=$(wavespeed-cli upload ./input.png)
wavespeed-cli run wavespeed-ai/seedance-v2 '{"image":"'"$URL"'","prompt":"camera dolly"}'

# Text → video w/ audio
wavespeed-cli run wavespeed-ai/veo-3 '{"prompt":"thunderstorm at dusk, cinematic"}'

# LLM
wavespeed-cli llm anthropic/claude-opus-4.6 "Summarize in one sentence."

Python SDK

import wavespeed
out = wavespeed.run("wavespeed-ai/z-image/turbo", {"prompt": "Cat"})
print(out["outputs"][0])

url = wavespeed.upload("/path/to/image.png")

Use the dedicated venv: ~/.local/share/wavespeed-skill/venv/bin/python script.py.

Errors

• 401 → key missing or revoked.
• 400 → payload doesn't match the model schema.
• 402 → insufficient balance / tier locked.
• 404 → task or model not found.
• 429 → rate-limited; back off.
• 5xx → upstream error; retry idempotent submits.

Refs

• Docs: https://wavespeed.ai/docs
• Skill repo: https://github.com/wesleysimplicio/WaveSpeedAI-Skills
• Python SDK: https://github.com/WaveSpeedAI/wavespeed-python
• API key: https://wavespeed.ai/accesskey

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: wesleysimplicio
• Source: wesleysimplicio/WaveSpeedAI-Skills
• License: MIT
• Homepage: https://wavespeed.ai

Install and usage instructions live in the source repository linked above.