Developer tools · DeepSeek Harness plugins — Page 125

古法编程模式（Classical Programming Mode）：AI 只出骨架、用户补全实现、AI 验收讲解的趣味编程练习模式。DeepSeek Harness 双面插件包（Host + Client）+ 随包 Agent 预设组合。

生成前注入人味风格约束 + AI 味检测报告 — 判断交给代码，风格交给 system prompt。

ide-ui host half: the `ide` Remote namespace (filesystem / git / search primitives) for the DeepSeek Harness Web UI. Unofficial community plugin.

Inline previews and clipboard copying for DeepSeek Harness read_image tool calls, including nested Code Mode calls.

DSH web plugin: set a background image and adapt the theme colors (background + text + brand + borders) to the image palette.

Load a workspace's own VSCode/Copilot AI configuration (.github/copilot-instructions.md, .github/instructions, .github/skills) into every DeepSeek Harness session. DSH 加载 VSCode AI 配置插件。

一键把输入框草稿改写为更清晰、更贴 Agent 执行的提示词：直接替换 + 一键撤回，带保真闸与长度闸。

Switch the agent preset (conversation mode: Standard / Code / Minimal / Creator) of a session mid-conversation, straight from the session header.

DSH 任务提醒插件：任务完成/需要批准/需要回答时发系统通知（可选页面浮层与自定义音频），三类事件独立开关

Claude Code style /init command for the DeepSeek Harness (DSH): generates a minimal CLAUDE.md and links AGENTS.md to it as a relative symlink.

See your markdown rendered live as you type in the DeepSeek Harness (dsh) web input box — headings, lists, tables, code, math and more, previewed right above the composer before you send.

Automatic J-Space cognition suite for DeepSeek Harness (dsh): a zero-config dsh-plugin that adds structured working state, checkpoints, and verify-before-ship discipline to every agent session.

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。

DSH bundle: jj (Jujutsu) version control via a zero-dependency MCP stdio server — exposes jj status/log/diff/describe/new/commit/edit/abandon/undo/rebase/fetch/push/bookmark as mcp__jj__* tools.

Job Hunting DSH plugin and runtime skill scaffold for local job intelligence workflows.

DSH plugin: decode and inspect JWT tokens (debug only, no signature verification).

Evidence-driven architecture reconnaissance for DeepSeek Harness

Codex CLI tool surface written in DSH form, distilled from @openai/codex v0.147.0 (tool handlers unchanged from v0.146.0).

Grok Build tool surface written in DSH form, distilled from xai-grok-tools.

Host-plane kernel mesh for DeepSeek Harness: kimi/grok/codex/minimax kernel adapters, L2 subagent recipes, and kernel tools.

Mini-Agent tool surface written in DSH form: read_file/write_file/edit_file/bash/bash_output/bash_kill/get_skill/list_skills/record_note/recall_notes. Schemas distilled from MiniMax-AI/Mini-Agent source.

DSH bundle (组合包): read-only project knowledge graph — modules, tech stack, architecture layers, file-level import graphs, observations. Provides /kgraph command, project_kgraph model tool and a browser UI panel.

DeepSeek Harness plugin for Kimi Code subscription OAuth

Kimi (Kimi Code account) integration bundle for DeepSeek Harness: provider route, models, device-code login and auto token refresh (npm: dsh-kimi-provider)