Plugins

A pre-execution guard for AI coding agents. It blocks destructive Git and file system commands, plus common attempts to access sensitive files, before a tool call runs. Supports Amp Code, Antigravity CLI, Claude Code, Codex, Cursor, DeepSeek Harness, Gemini CLI, GitHub Copilot CLI, Grok Build, Hermes Agent, Kimi Code, OpenClaw, OpenCode, and Pi.

White-box AGI architecture exploration: metacognition (self-cognition loop), continual learning (knowledge flywheel), world model (condition space, spatiotemporal memory graph), self-improvement (bootstrap discipline), zero-LLM white-box pipeline, and auditable trust guardrails.

Second-model auto-review on the approval answerer chain: a read-only reviewer subagent returns structured allow/deny verdicts with reasons, fail-closed by default.

GitHub-topic-driven plugin & skill marketplace: a Settings page that browses the auto-collected registry (the whole dsh-plugin topic plus the skills index, CI-refreshed every 2 hours) with one-click install, type detection, install-script and host-shadow-dependency safety confirmations, env-key management, and the STANDARD.md recognition spec.

Rewind conversation and workspace state, powered by a persistent Change Ledger.

Profile-wide DSH governance and routing with a Web Control Center for responsibility and evidence inspection, plus compaction, scoped semantic memory, safety continuity, and verified delivery; compatible with DSH 0.1.5-rc.1.

Claude Code-style declarative permission rules: ordered allow/deny/ask YAML rules matching tool names, arguments, workspace paths, and agent identity on the tools/pre-execute waterfall, with full session-log audit, dry-run mode, and hot reload.

In-place conversation rewind in the same session window without forking (Claude Code /rewind semantics): a per-message ↶ button cuts the model context back to any user message, with an optional Claude-Code-style file restore from disk-persisted before-backups.

Framework-upgrade safety and plugin gating for DSH: a pre-upgrade session-format contract preflight that adapts incompatible plugins and agent presets before you upgrade, automatic rollback on failure, auto-quarantine of plugins that broke a boot, plus environment fingerprinting and a public contract-rule pack. The built-in multi-source marketplace is a discovery layer only.

Commit-time audit harness for AI coding agents: 24 git-diff rules (secrets, out-of-scope edits, prompt injection), HMAC-signed audit trail, snapshot rollback, and an MCP server with 84 tools. Installable via dsh plugin add.

Feishu/Lark bridge for DeepSeek Harness: scan-to-connect PersonalAgent binding, streaming cards, git-worktree project workspaces, parallel per-scope tasks, multi-role agents, cross-session notify, in-chat model/key management, and a safety-net guardian that still answers in Feishu after dsh crashes.

Cross-platform Computer Use for DeepSeek Harness: virtual-mouse human-like operation (screen_observe + computer_click + type + key + scroll + drag, 11 tools), AX-tree zero-vision-cost mode with free GLM-4V-Flash vision fallback, safety guards (snapshot TTL / app whitelist / dangerous-action approval / password-box protection).

Auto mode for DeepSeek Harness (dsh): an Auto entry in the Permissions selector whose approval prompts are answered by OpenGuardrails policy instead of a human — plus the full OGR guard engine underneath. No core changes.

Classifies pending tool calls with Jev (TypeSafe AI's non-generative decision model) on tools/pre-execute — confident high-risk calls are denied, ambiguous ones escalated to approval — and auto-approves clearly-granted reversible calls on approval/request behind argument-evidence gating. Also subclasses the session-reference resolver so snapshots keep Jev-scored messages instead of dropping oldest-first. Degrades to stock behavior on any provider failure; shadow mode with a /jev-stats command; TypeSafe or OpenRouter endpoints; 64 tests.

Install safety net for DeepSeek Harness: pre-install snapshots, one-click/automatic rollback, guarded boot, and incident reports that auto-trigger agent analysis. 中文: DeepSeek Harness 插件安装安全网（安装前自动快照、一键/自动回退、守护启动、事故报告自动触发 Agent 分析）。

Claude Code /rewind for DeepSeek Harness: git-first workspace snapshots before every mutating tool execution, turn-boundary session forks, and a one-shot /rewind command that restores files and forks the session back to a checkpoint.

Harness-neutral reliability infrastructure for AI coding agents: reversible destructive actions, pre-emission redaction, recovery evidence, and a shared Decision Protocol with optional native adapters.

Drive OriginLab Origin scientific plotting from AI chat via MCP: 62 tools. Modal-dialog watchdog, LabTalk safety gate, release/reconnect, fit with initial/fixed/weighted params, fine-grained plot and sheet edits, matrix tools, FigureSpec declarative YAML (diff-able and replayable), MCP Resources read-only snapshot, matplotlib-figure bridge, PowerPoint assembly with panel labels, Graph Gallery template search, EPS export, file-access whitelist, visual regression baseline, and editable OPJU delivery with deterministic readback verification.

Dispatch subagents with named profiles (model, reasoning effort, and tool scope per task), safety checks, cost estimation with savings comparison, and a full dispatch decision ledger.

Enhanced vision toolbox: 14 pixel-level vision tools (describe, ground, detect, crop, pixel-diff, OCR, long-screenshot OCR, vectorize, colors, cutout, screenshot, present, materialize, html-screenshot) driven by one OpenAI-compatible endpoint, with clean \[图片: path] bridge markers, content-safety classification and rate-limit auto-retry.

Host-side guard that lets the write tool only create new files: when the target already exists it denies the call before execution and tells the model to use edit instead, applied to all sessions.

Adds session, project, and global controls in the DSH WebUI for skills, MCP servers, tools, prompt injections, approval escalation, and safety guards; enforcement is scoped to the current agent.

Adds persistent execution contracts, recursive work graphs, critical clarification, and evidence gates for long or underspecified Harness tasks.

Ask-before-act approval for every edit and every bash command: a red/green line diff before a file is touched, a description+command panel before a command runs — approve once or reject, each gate with its own master switch in Settings → General.