screenshot-review
Run the following command in DeepSeek Harness:
dsh plugin install wenyixiaoqingnian/screenshot-review
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install wenyixiaoqingnian/screenshot-review in the DeepSeek Harness terminal to install the plugin; the source repository is https://github.com/wenyixiaoqingnian/screenshot-review — restart dsh web after installation to activate the skill.
About this plugin
When using an LLM to write frontend code, the most frustrating step is rarely generation itself — it is verification. The model can only see the raw HTML and CSS it produces, not how the browser actually renders them, so every visual fix depends on a human saying "make it look better" and hoping for the best.
screenshot-review closes that gap. It gives an image-capable model a self-directed review loop: open the page, take a screenshot, read that screenshot back, assess the visual result, revise the code, and screenshot again. The iteration runs automatically until the rendered appearance converges on the user's intent or a configurable round limit is reached — no manual frame-by-frame comparison needed.
If you rely on AI to generate or refine UIs (landing pages, component libraries, dashboards, etc.) and your model supports image input (deepseek-v4-flash, GLM-4V, and similar), this skill turns the "screenshot → feedback → fix" cycle into a hands-free loop, cutting down the back-and-forth with a human reviewer.
Use Cases
- Generate a landing page with AI, then let the model auto-screenshot and iterate until visuals meet the target
- Refine component styles by having the model screenshot and self-review, reducing manual visual comparison
- Bulk-generate UI pages and use the screenshot loop to verify rendered output matches design intent
Best For
- Engineers who use multimodal LLMs to assist with frontend development
- Developers who frequently generate or refine UI code with AI and want fewer manual feedback rounds
- Automation teams building DSH workflows who want to add a visual verification step
Related Plugins
A method pack that makes AI coding agents plan against your real baseline, prove completion with fresh evidence, and reduce reworks and unsafe changes.
Turns the DeepSeek Harness session into a captain that builds a durable sub-agent team, splits goals into dependency-aware tasks, and coordinates work via direct messages and a live Web UI.
Gives coding agents design judgment, letting Claude Code, Cursor, and 70+ agents generate and iterate high-quality UI, presentations, and graphics right from the terminal.
Run the Pi ecosystem's plugins on DeepSeek Harness, unmodified, via a compatibility layer that implements Pi's public extension ABI on DSH's native services.