Best Image Generation

**Judgement:** This skill is a capable, well-documented integration for high-quality AI image generation and editing via the EvoLink API, suitable for many production-style visual workflows where external API dependence and content-safety concerns are acceptable. **What it does well** - Covers core imaging modes: text-to-image, image-to-image, and editing in one interface. - Supports a good range of aspect ratios and quality levels (1K–4K), allowing control over output fidelity vs. cost. - Implements a clear asynchronous pattern (submit + poll) with extended retry logic, improving robustness for longer generations. - Simple natural-language triggers in English and Chinese lower friction for end users. - Built-in local download and `MEDIA:<path>` output make it straightforward to attach results inside OpenClaw. **Key risks / limitations** - **External dependency & cost:** Relies fully on the EvoLink API and a valid `EVOLINK_API_KEY`; outages, latency, rate limits, or pricing changes at EvoLink will directly affect this skill. Higher quality (4K) explicitly incurs extra charges. - **Latency:** Polling every 10 seconds for up to ~12 minutes can introduce noticeable delays for complex requests; not ideal for highly latency-sensitive interactions. - **Content safety:** Documentation does not mention any safety filters, moderation, or guardrails for harmful, NSFW, or copyrighted content, so it should be treated as minimally moderated and paired with external policy checks if used in regulated settings. - **Error handling details:** While status polling for `completed`/`failed` is defined, the page does not describe detailed error reporting, backoff strategies, or graceful degradation under partial failures. **Recommended scenarios** - Best suited for assistants or tools that need **high-quality illustrative or creative images** on demand (marketing, content creation, prototyping) and can tolerate external API latency and cost. - Appropriate for **multilingual user bases** (English/Chinese) needing simple, prompt-based image generation or editing with minimal configuration. - Less suited for **highly safety-sensitive** or **strictly on-premise** deployments unless combined with separate moderation and reliability layers.