Design-fidelity QA for DeepSeek Harness: lend any text-only model an eye, then judge whether the implementation matches the mock. Ships the benchmark behind that judgement — four fixtures, 23 injected defects, and every raw model transcript. Retires itself when DeepSeek ships vision.
DeepSeek Harness 自动审批插件:在 workspace-write 沙箱之上叠加确定性规则 + LLM 安全审批,自动模式不放宽沙箱、fail-closed。 Safe auto-approval for DeepSeek Harness — deterministic rules + LLM review on top of the workspace-write sandbox. Auto mode without ever granting full-access.
DeepSeek Harness all-in-one: no model switching — regular DeepSeek auto-routes to vision & image gen. Multi-backend: Gemini + any OpenAI-compatible (GPT-4o, Qwen-VL, GLM-4V, gpt-image, DALL-E, Flux, OpenRouter). gemini_vision/gemini_generate_image/gemini_optimize_image with vision self-check. Better than modlens.
DeepSeek Harness (DSH) plugins. qwen-image gives a text-only coding model eyes: an image goes to a Qwen-VL route through ctx.llm and comes back as text, so DeepSeek keeps coding while Qwen looks. Pure ESM, no build permission at install. | DSH 插件集:qwen-image 让纯文本模型借千问 VL 读图,返回文本;纯 ESM,安装无需构建授权。
DeepSeek Harness dynamic plugin: a round tidy button left of the model selector — click tidies the working folder, long-press picks another folder. 整理文件夹按钮插件
Persistent, model-editable memory/notes store for DeepSeek Harness. Adds memory_set/get/delete/search tools backed by ctx.storageDomain so facts survive across sessions (dsh-plugin).
DSH Web GUI plugin: session-scoped "Allow for this session" option in the approval dialog (per-mode standing grants, localStorage) + right-panel auto-open guard
为 DeepSeek Harness 提供类似 Codex Auto Reviewer / "approve for me" 的权限模式。A permission mode for DeepSeek Harness similar to Codex Auto Reviewer / "approve for me"
DSH-Plugin: skill trigger/selection gate - user picks from described candidates when the model triggers a skill; 10s no-choice falls back to the system default.
The auto permission tier for DeepSeek Harness: workspace-write requests run directly, sandbox escalations (danger-full-access) are judged by a separate classifier model.
OwlX MCP server - live crypto structure scores, MemeSniper, token financials and recorded signal hit-rate as MCP tools for Claude, Cursor, DeepSeek Harness, Grok and any MCP client.
DSH 中文增强补全:斜杠命令名与插件清单卡片中文化,支持三种显示模式(只显示中文/中英都显示/不翻译)。Chinese-first plugin for DSH: translates slash-command names and plugin-inventory cards, with three display modes.