Plugin

Verification gate for DSH agents: every acceptance criterion must be backed by server-stamped real tool evidence before the completion gate lets a goal through (advisory audit, enforce gate, durable permits).

Human-in-the-loop tools (`human_task` / `human_task_ready_check`): pause the agent until the user performs a real-world action or returns an observation (GUI operation, game testing, visual verification, hardware checks), gated by session consent and an AFK presence check, returning a structured JSON result via a web task dialog.

An MCP lazy-loading, dynamic tool-loading, and Tool Router plugin for DSH. It addresses context bloat, wasted tokens, oversized tool catalogs, and persistent tool schemas when many MCP servers are configured. The plugin searches, routes, and loads MCP tools on demand only when the current task needs them, then unloads idle tool schemas at turn end to reduce context usage and save tokens while keeping connections warm for fast reuse. Supports explicit activation, automatic routing, paginated tool-catalog refresh, bounded reconnection, and both stdio and Streamable HTTP MCP transports.

DSH bundle that registers 13 native codegraph_* tools wrapping the codegraph CLI, so agents bootstrap, maintain, and query a pre-indexed code knowledge graph (symbol query, explore, node trail, callers/callees, impact, affected tests) instead of grepping files.

Per-model proxy routing for LLM requests: selected models go through a proxy (Clash etc.) while the rest stay direct, with automatic retry on transport errors, 429 and 5xx, configured live from the settings page.

Tiered model routing for DeepSeek Harness: think/build model tiers, vision delegation (the vision model only describes images; the working model acts), image generation, and a high-impact guard.

Turn histogram navigation rail for DeepSeek Harness Web with mixed, DeepSeek, and Codex visual variants.

Rotates the model status line through a 53-phrase English/Chinese deep-theme pool: a random pick at turn start, event-driven switching on new thinking segments or tool calls (3s throttle), and a click on the status line toggles the language.

Auto-rewrites the composer draft into a clearer, more executable prompt after you stop typing, using scene-aware playbooks, preserving long attachments verbatim, and numbering multi-part questions.

Left-edge conversation rail for the Web UI: one bar per turn, length scaled to the turn text, hover for a question and answer preview, click to jump; bars come from a host-side outline of the whole session log, so turns outside the loaded window are drawn too and history is paged in on click.

Composer GitHub picker: click the GitHub icon at the bottom-right of the input box to search the workspace repository's issues and pull requests via the gh CLI, and insert the reference as a URL or @owner/repo#number mention.

Add six languages (繁體中文, 日本語, 한국어, Français, Deutsch, Español) to the DeepSeek Harness Web UI language options, with hand-polished per-language translations and automatic fallbacks (Simplified-to-Traditional conversion for zh-TW, English for the others) covering upstream updates and third-party plugins.

Code knowledge graph for dsh: indexes symbols, call sites and imports into SQLite and answers call/dependency questions via CLI or stdio MCP tool server.

Decodes a JWT for debugging without verifying its signature, and flags expiry.

Exports the current session as a deterministic Markdown handoff document.

Formats and validates JSON/YAML/TOML/SQL, with line-and-column error locations.

A-share and US stock trading toolkit for DSH agents: realtime quotes, OHLCV klines, ADX regime signals and simple backtest previews via EastMoney. Read-only, never places orders.

One-click Windows/WSL2 environment switcher for DSH (coexistence mode): both environments run simultaneously on separate ports and switch from the web UI without killing any process.

Codex-style token usage dashboard: five stat cards, GitHub-style activity heatmap (daily/weekly views), insights & model ranking; persisted snapshot with incremental sync, survives session deletion.

System parameter adjustment UI: grouped settings page for timeouts, parallelism, LLM retry policy, model params, web search, permissions and presets; safe profile-config editing with line-level YAML matching and post-write validation.

Chinese localization for tool results: intercepts tool execution output and translates English markers (`[exit code]`, `[timed out]`, `[sandbox: ...]`, etc.) into Chinese, with a settings toggle.

Splits the composer model selector into a searchable, favorite-marking dropdown and a reasoning-effort slider, with Ctrl+P / Ctrl+T quick-switch shortcuts.

Model-agnostic visual-inspection pipeline for text-only agents: page-by-page HTML screenshots plus a ready-made vision-subagent briefing contract (gewu_prep), then source-code truth verification of every finding (gewu_locate); validated on mimo-v2.5 & qwen3.7-plus.

用于DSH的对话框任务音效，以及个性化聊天背景||Dialog box sound effects for DSH, as well as personalized chat backgrounds.