メモリ・ナレッジ · DeepSeek Harness プラグイン — 10 ページ

Thinking-chain memory for DeepSeek Harness (all-sessions edition): remember WHY a decision was made, recall it by keyword or meaning, inject it into every session's prompt, and browse it as a 3D nebula graph.

Claude Code-style @path imports for DSH: expands @imports in AGENTS.md/CLAUDE.md and injects the referenced files into model context at session start

AI deep-reading assistant: extract core claims, argument structure, and key evidence from books and articles. Five modes (quick, deep, knowledge map, Feynman, full book); knowledge-map reports include confidence levels, data facts, relation labels, Mermaid mindmaps, and recall questions. Reads WeChat article URLs, TXT/Markdown/HTML/PDF files, or pasted text (built-in pure-JS PDF extractor). Exports Markdown, FreeMind/XMind mindmaps, and HTML reports. DeepSeek Harness plugin.

知识库驱动的 Agent 预设

Native Graphify knowledge graph plugin for DeepSeek Harness (DSH)

Obsidian side-panel math memory agent for DeepSeek Harness: layered long-term memory, idea memo with proactive review reminders, unified note_recall retrieval plus dedicated note search/create/backlink tools. Ships as both a dsh plugin and an Obsidian community plugin.

cross-session memory for DeepSeek Harness v0.10.0：记忆层（触发式召回 · 类型优先级 · 五条清理规则 · 工作流链 · 纯本地单文件 SQLite）。

同事引擎 for MyWork Kit：每位同事一条永远的对话（自己的 dsh 会话、文件夹、例行与记忆），发一句话就干活，交付文件并在后台核验。同事页与文件页，deliver / mywork_ask / mywork_routine_create 等工具，场景插件通过 myworkTasks 服务注册。

Obsidian Vault Studio for DeepSeek Harness: local vault index, hybrid semantic search, and safe note writing.

DeepSeek Harness host plugin that keeps project documents and long-term memory as plain Markdown in a dedicated Obsidian vault.

Project board, task schedule (gantt) and project memory for DeepSeek Harness: sessions opened from a card auto-load it.

DSH web plugin: terminal-style input for the composer — bash-like Up/Down prompt history with prefix search and Ctrl+R, selection toolbar with copy and quote (Codex-style > blockquote), right-click paste, toggleable draggable chat TOC, fully internationalized (zh/en), with copy modes in Settings (history sourced from the session's own message log; zero config required).

dsh plugin: git-like conversation forking (branch refs on sessions)

把 TiddlyWiki 5 变成你和 Agent 共用的本地文档中心（纯文本 + git）：你在 DSH 界面里阅读、写作、记日记，Agent 用 16 个 tiddlywiki_* 工具读写同一批笔记；每条笔记都是你能读、能改、能审计、能随时带走的文件——不是自动记忆插件，不做后台抓取。可选一键发布到微信公众号草稿箱。 Turn TiddlyWiki 5 into a document center you and your agent share, in plain text under git — a knowledge base you can read, edit, audit and take with you, not an auto-memory plugin. Mounts via the official dsh plugin system; no DSH source changes.

Evidence-driven learning memory and automation for Git-native AI agents

AI 陪伴技能：人设配置、长时记忆、情绪识别与反馈、多轮情感对话

Host-only Obsidian integration for DeepSeek Harness: 25 on-demand vault, knowledge, attachment, link, editor, and command tools with an authenticated local IPC companion.

DSH bundle plugin: mounts Project Memory skills, bootstraps memory architecture on first use, registers the /project-memory slash command, and prompts for memory compounding after tasks.

Unofficial DSH plugin: multi-session collaboration with work orders, handoff, and file-based memory.

Official hybrid DSH plugin: Cheat Engine dynamic debugging toolkit with host ce_* tools and a client floating status panel.

LLM Wiki panel: right-column file tree, knowledge graph, markdown/PDF reader, and upload flow for the DeepSeek Harness wiki

QQ channel plugin for DeepSeek Harness via OneBot v11, with a bundled Life Engine component (memory plasticity, inner monologue, sleep consolidation, KB search)

Learning assistant extension for DeepSeek Harness (dsh): DeepTutor tutoring for your agent — deep explanations, self-test questions, learning paths, personal knowledge-base search (RAG), and note archiving (HTTP/WS first, CLI fallback; auto-adapts local/remote deployment).

Self-evolving pitfall immunity for coding agents (Claude Code, Codex, Cursor, ZCode, DSH): capture failed-then-fixed tool episodes as SQLite+FTS5 antibodies and inject them before the agent repeats the mistake. Zero runtime dependencies.