技能详情（站内镜像，无评论）
作者：Ilya @al1enjesus
许可证：MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本：v1.1.0
统计：⭐ 0 · 489 · 2 current installs · 2 all-time installs
⭐ 0
安装量（当前） 2
🛡 VirusTotal ：可疑 · OpenClaw ：可疑
Package：al1enjesus/wavespeed
安全扫描（ClawHub）
- VirusTotal ：可疑
- OpenClaw ：可疑
OpenClaw 评估
The skill broadly matches an image/video-generation tool, but it inconsistently hides that it requires an API key and contains a few instruction/packaging mismatches that could lead to accidental exposure of secrets or sloppy installs.
目的
The skill's stated purpose (WaveSpeed image/video generation) matches the code and model list: the CLI talks to api.wavespeed.ai and exposes model aliases. However the registry metadata says no required environment variables or primary credential, while both SKILL.md and the script clearly require WAVESPEED_API_KEY. That discrepancy is incoherent and important: a user installing this skill would not be warned that a secret is needed.
说明范围
SKILL.md instructs the agent to check the WAVESPEED_API_KEY env var (and even suggests running echo $WAVESPEED_API_KEY). Asking the agent/user to echo an API key risks accidental leakage into logs or chat. The instructions also say to check TOOLS.md and to 'ask the user' if no key is found; those are reasonable, but the explicit echo advice is risky and unnecessary for normal operation.
安装机制
There is no install spec (instruction-only) which lowers install risk. Minor packaging inconsistencies: README suggests installing 'axios form-data' but the shipped script only uses built-in https/fs and package.json lists no dependencies. This looks like sloppy packaging rather than active malicious behavior.
证书
The code requires WAVESPEED_API_KEY (process.env.WAVESPEED_API_KEY) and will exit if it's not set, yet the skill metadata declares no required env vars or primary credential. The SKILL.md also asserts the key is 'already set in all Clawster containers' — an unverifiable and suspicious claim. The instruction to echo the env var could expose the secret; environment access is more privileged than the metadata indicates.
持久
The skill does not request always:true, does not modify other skills, and is user-invocable. It does not attempt to persist itself or change system-wide settings. No elevated persistence privileges are requested.
安装（复制给龙虾 AI）
将下方整段复制到龙虾中文库对话中，由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库（龙虾 AI）执行：为本机安装 OpenClaw 技能「WaveSpeed AI」。简介：Generate and edit images and videos using WaveSpeed AI's 700+ model library. Us…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装：https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/al1enjesus/wavespeed/SKILL.md
（来源：yingzhi8.cn 技能库）SKILL.md
---
name: wavespeed
description: "Generate and edit images and videos using WaveSpeed AI's 700+ model library. Use when the user wants to generate images from text prompts (FLUX, Seedream, Qwen), edit or retouch photos (nano-banana-pro/edit keeps faces identical while changing clothes/background), generate videos from images or text (Kling, Veo, Sora, Wan, Hailuo), or upscale videos to 4K. Triggers include: generate an image, edit this photo, change the background, swap the outfit, make a video from this image, animate this photo, upscale this video, use WaveSpeed, use nano banana pro."
---
# WaveSpeed AI
700+ AI models (Google, OpenAI, ByteDance, Kling, Luma) via one API. Images in <2s, videos in <2min.
## API Key — check in this order
1. **`WAVESPEED_API_KEY` env var** — already set in all Clawster containers, just use it directly
2. `TOOLS.md` in the workspace — look for `WaveSpeed AI` section
3. Ask the user
**Never search for the key** — if `WAVESPEED_API_KEY` is in the environment, it's ready to go. Check with:
```bash
echo $WAVESPEED_API_KEY
```
### Get your API key
Sign up at **[wavespeed.ai](https://wavespeedai.pxf.io/3kPoRd)** → Dashboard → API Keys.
New accounts get free credits. Pay-as-you-go pricing — no subscription required.
```bash
export WAVESPEED_API_KEY=your_key_here
```
The skill script is at `skills/wavespeed/scripts/wavespeed.js`.
## Usage
```bash
# Image generation
node wavespeed.js generate --model flux --prompt "sunset over mountains" --output out.png
node wavespeed.js generate --model seedream --prompt "..." --size 1024x1024
# Image editing (face/portrait-safe — preserves identity)
node wavespeed.js edit --model nbp --prompt "change bathrobe to black hoodie, dark background"
--image https://example.com/photo.jpg --output result.png
# Video from image
node wavespeed.js video --model wan-i2v --prompt "slow cinematic zoom"
--image https://example.com/frame.jpg --output clip.mp4
# List all aliases
node wavespeed.js models
# Check task status
node wavespeed.js status --id task_abc123
```
## Key Models (Quick Reference)
| Task | Alias | Best for |
|------|-------|---------|
| Edit photo keeping face | `nbp` | Portrait retouching, outfit/bg change |
| Fast image gen | `flux-schnell` | Drafts, quick tests |
| Best image quality | `flux-pro` / `seedream` | Final outputs |
| Image → Video | `wan-i2v` | Fast, affordable |
| Premium video | `kling` / `veo` | Cinematic quality |
| Text → Video | `sora` / `veo` | Story videos |
See `references/models.md` for full model list with IDs, params, and pricing.
## Important Notes
- **Image editing** (`nbp`, `nb-edit`): always pass images as `images: [url]` array — this is required
- **Face preservation**: `google/nano-banana-pro/edit` is the best model for editing photos while keeping the person's face identical
- Output files are saved to current directory by default; use `--output` to specify path
- Videos can take 2-5 minutes; script auto-polls with progress indicator
- For multiple input images (multi-reference editing), use `--images url1,url2`