Agent collaboration · DeepSeek Harness plugins — Page 308

收集当前 Agent 可见的本地 Skill、原生 Tool 与 MCP Tool，提供必要的可视化编排和约束编辑能力，并导出稳定的 YAML 中间表示。后续可由大模型及对应框架把 YAML 生成 Python、`workflow.mjs` 或其他代码工作流。

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。

Git worktree isolation backend for DeepSeek Harness workflows: binds each write-capable workflow agent to its own recoverable checkout.

Project-scoped organization and long-lived AI workforce layer for DeepSeek Harness

DSH 插件：把 workspace 内容与 DSH agent 任务能力暴露为 HTTP API（/workspace-api/*），其他程序可查询目录、搜索文件、提交自然语言任务给 agent 处理。

DSH Web plugin: Archive-All action on every workspace folder menu in the sidebar (Ungrouped bucket included) — archives all unarchived sessions under that folder after one confirm; running, currently open, blank and subagent sessions are skipped.

DeepSeek Harness plugin that surfaces concurrent workspace changes before an agent continues.

DSH web plugin: an IDE-style Changes panel in the right Sidebar — grouped working-tree changes (JetBrains-style Changes/Unversioned groups, directory tree) with a full diff viewer (unified/side-by-side, line numbers, file navigation), copy-diff, and open-in-IDE. Read-only against the repository.

Out-of-tree DSH bundle: applies the local machine's native direnv environment to explicitly Agent/workspace-owned Bash executions and local stdio workspace MCP rows. Depends on dsh-workspace-overlay; requires a host-installed direnv; never calls allow/deny or reads .envrc itself.

Approval-gated workspace instructions and memory for DeepSeek Harness

Per-workspace Agent preset bindings for DeepSeek Harness: bind a preset to a workspace once and every new session there boots with it.

Per-workspace custom system prompts for DeepSeek Harness (DSH): set one active prompt per workspace, captured into a session at creation time, editable from a session-header dialog. 给 DeepSeek Harness 的「工作区自定义提示词」插件：每个工作区一条生效提示词，新建会话时固化注入，会话头部弹窗增删改查。

A fourth permission option for DeepSeek Harness sessions: selected-workspace-write, where the user opens the workspace directory tree and toggles which directories the agent may write to.

Read-only Git worktree discovery, status, and comparison tools for DeepSeek Harness

Per-task Git worktrees for DeepSeek Harness: a task space spans one or more repositories, is registered as a DSH Workspace, and finishing merges the branches, removes the worktrees and files the documents.

DeepSeek Harness plugin that ships the world-engine and world-agent skills: compile a described world into declarative rules, run multi-agent scenes, and let agents, subjectivity, and emotion emerge instead of being encoded.

Run the DeepSeek Harness agent's filesystem, bash and PTY terminals inside a Docker container, so execution state survives a harness restart.

A safe, bounded, read-only perception layer for DeepSeek Harness agents: observe administrator-defined JSON APIs with provenance, snapshots and deterministic diffs — never an arbitrary HTTP client.

Commitment write-gate for AI coding agents: two-tier (deterministic + LLM judge) pre-execution policy. Engine-agnostic core with a DeepSeek Harness (dsh) adapter.

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。

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。

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。

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。

WSL execution for DeepSeek Harness on Windows: ships the minimal-wsl agent preset (bash executes inside WSL via wsl.exe, non-persistent + persistent PTY), a general-purpose wsl model tool for standard/cordis presets, and a Web Plugins settings tab with distro diagnostics over the /wsl-manager JSON API.