Agent 連携 · DeepSeek Harness プラグイン — 286 ページ

DeepSeek Harness bundle: two academic research agent presets — scopus-scholar (搜索驱动模式, Scopus/OpenAlex search with a claim-verify loop) and scopus-knowledge (知识驱动模式, a code-enforced knowledge pipeline consulting a tool-free 科学顾问 persona)

Bundles the official Scrapling web-scraping agent skill (anti-bot bypass, stealth browsing, spiders, adaptive parsing) into a DeepSeek Harness profile.

Desktop & web UI automation for DSH: full-screen capture, Windows OCR, mouse click/drag, typing, key press, window activate, plus battle-tested Paint drawing capability

DeepSeek Harness (dsh) plugin that bridges the local 'Screen Automation Helper' desktop platform (Windows/macOS) into the Agent tool system. Exposes status, capabilities, workflow list/run/stop, run management, screen capture, local recognition primitives, input control, workflow lifecycle, task sessions and VLM understanding as dsh tools that shell out to the helper's CLI (JSON interface).

DSH 桌面版输入框工具条插件：在「听写」左边加「录屏」与「截图」两个按钮 —— 截图带选区，录屏产出一个 .webm 视频并自动进附件。Screen capture & recording buttons for the DSH desktop composer.

Autonomous screen vision for DeepSeek Harness on macOS and Windows: the agent captures the screen and receives the image itself in one call.

Give the agent eyes on the screen: capture the whole desktop, one window, or one region and hand the IMAGE ITSELF to the session model (no second model, no description layer), plus an exact local pixel diff that locates what actually changed between two images instead of asking a model to guess. Windows only.

截图审阅 dsh skill: 模型自己截图、自己看图、自己改代码，迭代优化前端效果

DSH 插件：在 write/edit 缝上拦住踩过的 shell / PowerShell 坑（set -e 命令替换陷阱、递归 grep 裸变量、.ps1 缺 BOM）

Agent-preset multi-channel web search tools and profile settings for DeepSeek Harness.

深度搜索插件：三层检索（表层多引擎/深网挖掘/Tor代理）+ 23 工具（搜索/抓取/OSINT/归档/分享检索）

Agent-native search, log investigation and cluster operations for DeepSeek Harness. OpenSearch-first, provider-neutral by design.

SearXNG-backed search provider for the DeepSeek Harness web capability seam (ctx.web)

SQL 优化模式：EXPLAIN 分析、索引策略、N+1 解决、查询改写。受 wshobson/agents（38k★ MIT）启发。

安全加固：威胁建模（STRIDE）+ 三层边界系统，任何接收用户输入/鉴权/存储敏感数据/集成外部服务的功能都按安全第一开发。受 addyosmani/agent-skills（88k★ MIT）启发。

开眼看世界：DSH 决策层插件——让 agent 在「该上网搜索」的时候先搜再答（fresh-info-first / 联网优先）。判定『何时搜索』（决策层），不自带搜索引擎（能力层）。

Yogacara eight-consciousness agent society for DeepSeek Harness: auditable double-loop runtime bridge, mneme memory consolidation tuning, MCP society tools, and six seed skills

Frozen, topology-aware session maps for DeepSeek Harness

DSH plugin: select text in any Windows app to explain or translate it with DeepSeek Harness's own agent, answered in a native always-on-top overlay (draggable, resizable). · DeepSeek Harness 划词工具：任意应用里选中文字即在锚点右下角弹出菜单，用 Harness 自己的 agent 解释/翻译（自动带上选区所在的上下文段落），结果落在零依赖的原生置顶浮窗里（可拖动/缩放）。

Self-evolution for DeepSeek Harness: bounded self-managed memory, an automatic per-turn review that writes plain-language notes and refines skills instead of duplicating them, a skill health check, cross-session recall, a durable learning row inside the conversation itself, and a card above the composer reserved for changes waiting on your decision.

Self-evolving plugin for DeepSeek Harness: observes agent behavior, proposes improvements to its own skills/policies via LLM, validates them in a sandboxed trial agent, and applies versioned mutations with rollback.

DeepSeek Harness 长期记忆与自进化插件：L0 对话捕获 → L1 记忆提取 → L2 场景归纳 → L3 用户画像，自动召回注入 + 技能合成，纯本地。

DSH 自记忆插件 v3（P0 升级）：在 v2（加权关键词/泛化/倒排索引/前提感知/同义词规范化）之上增加事件驱动自动采集（会话轨迹落盘）、pre-step 自动召回注入（<selfmem_memory> 围栏）、写路径可靠队列（失败入 pending 重放）、CJK 感知 token 预算。

DSH 自记忆插件 v2：加权关键词关联度、问题泛化（隐去细节）、倒排索引快速检索、前提条件感知（矛盾方案区分）、关键词规范统一。记录问题、多种解决方法、脚本文件与模型调用全过程，新会话先检索再执行再回写，支持外部导入。