Agent 連携 · DeepSeek Harness プラグイン — 147 ページ

Evidence-backed, human-approved project memory for DeepSeek Harness

Project Context Engineering for AI agents: turn messy project context into a clear, actionable workspace.

Private, per-project system-prompt rules for DeepSeek Harness — matched by git remote URL, repo path, or cwd prefix; worktree-aware; never committed to the repository.

PRO-LONG-style programmatic memory for DeepSeek Harness: appends every session event to a per-workspace log.txt that the agent retrieves with grep/python. Built-in write probe, /prolong status command and permission-deny counter.

Privacy-conscious Prometheus metrics and a Grafana dashboard for DeepSeek Harness

全局提示注入 / 上下文病毒（context virus）防御插件：tools/pre-execute 扫工具参数、tools/post-execute 扫工具结果、agent/pre-step 扫进入模型前的消息并注入金丝雀守卫、llm/stream 出站消毒；支持 block / quarantine / monitor 三模式、磁盘化可演进病毒库（学习/导入/导出）与审计日志。纯 JS 无原生依赖，profile 层挂载，对所有预设与子代理全局生效。

DSH 提示词工作台：看清 system prompt 的每一段、写下你自己的家规、体检重复与矛盾，并记下每次改动。

给 DSH 对话输入框加一个 ♂harden 按钮：一键把草稿锻造成硬邦邦、雷霆炫酷、肌肉团队味的提示词。A hard-boiled prompt rewriter for the DeepSeek Harness composer.

DSH 持久化 Web 插件（prompt-optimizer）：在输入框模型选择器左侧提供「优化提示词」按钮——用当前所选模型把模糊需求改写成可执行的优质提示词；优化期间输入框呈现跑马彩虹渐变边框并锁定输入，输入框上方显示优化状态。

Independent read-only acceptance layer for DeepSeek Harness: spawns a read-only verifier before each top-level turn closes and steers gaps back into the agent.

GitHub research assistant for DeepSeek Harness: model-facing tools to search repositories, read READMEs/source, and clone candidates, so the agent can prospect open-source solutions for an idea and synthesize an implementation plan.

A session-bound agent-native REPL for DeepSeek Harness PTC mode.

Phone-first PWA for DeepSeek Harness: mounts at /m on dsh web and talks to the same /api as the desktop GUI — sessions, streaming chat, approvals, agent questions. iOS + Android, zero install on the phone.

CodeAct agent loop for the DeepSeek Harness: a persistent IPython session as the model's action space, with harness tools bridged in as a virtual `__dsh__.tools` module.

Tencent QQ Agent Mail integration for DeepSeek Harness with a Service, event stream, Web settings, OAuth, and 14 structured tools.

QQ ↔ DeepSeek Harness agent bridge over OneBot v11 (reverse WebSocket): per-group/per-user session grouping, @-quote voice speech-to-text (GLM-ASR / OpenAI-compatible), quote & image resolution, face/sticker tools, /new /status commands. Independent bundle; remove anytime: dsh plugin --profile web remove dsh-qq-onebot-bridge

System actions for DeepSeek Harness agents: open URLs in the default browser and reveal files in the file manager (host-only, two model tools).

AI response quality review for DeepSeek Harness: audits each finished assistant turn with an independent reviewer model and steers the agent to fix unreasonable output (up to 2 review rounds).

提问面板折叠：agent 提问时把问答面板折叠成一条紧凑条（保留标题与取消），展开后回答草稿不丢。Collapse the DSH question panel into a compact strip; the unsent answer draft survives expand.

Event-Driven Autonomous Agent Runtime for DeepSeek Harness. Watch any event source (HTTP/file/command/webhook), match conditions, and spawn an isolated ReAct Agent Session that autonomously achieves your goal — with budget gates, retry, deterministic verification, failure classification, run trace, and a floating widget UI.

Real-time Computer Use for DeepSeek Harness on Windows: full-screen capture, timestamp replay, and mouse/keyboard injection

Web-profile helper that diagnoses and writes MCP config so Codex, Cursor, or Claude Code can spawn dsh --profile codex

Cross-session long-term memory for the DeepSeek Harness: a two-phase extraction/consolidation pipeline over a file-based memory workspace, recall injection, a citation feedback loop, lease-based job semantics, plus a settings page, a sidebar status panel and a memory viewer.

在 DSH(DeepSeek Harness)中创建定时提醒与重复任务:会话内即可设提醒,后台调度器到点触发。