@llangtop/dsh-approval-ai
AI approval answerer using the unified ctx.llm route with fail-closed local policy checks
191 results
AI approval answerer using the unified ctx.llm route with fail-closed local policy checks
DeepSeek Harness 自动审批插件:在 workspace-write 沙箱之上增加 Auto 档——确定性规则放行/拒绝,模糊操作由 LLM 裁决,危险操作转人工。保留工作区沙箱边界,不放宽为 full-access。
DSH Web GUI 审批语音提示:需要审批/回答的弹窗出现时,播放提示音并用语音播报,避免漏看
DeepSeek Harness plugin: per-edit approval with a red/green line diff before write/edit/str_replace_editor — approve once or reject (Claude Code behavior).
Cross-session memory with governance, self-evolution and an approval-gated write path for DeepSeek Harness. Heat-ranked recall, environment conventions, forgetting lifecycle, supersession chains, poisoning defense, auto-distillation, and an open measurement ABI.
Native, per-workspace long-term memory for DeepSeek Harness: facts and profiles on DSH's own storage-domain, cross-session recall through session-query FTS, approval-gated writes, and cited provenance — no external server, no extra runtime dependency.
DSH WebUI capability controls for skills, MCP servers, tools, prompt injections, approvals, and safety guards at session, project, and global levels.
Durable batch review and commit safety for DeepSeek Harness tool side effects
记忆搬家:把 Claude / ChatGPT / Claude Code 的历史记忆搬进 DeepSeek Harness —— 带逐字证据,逐条批准 | Move your Claude & ChatGPT memory into DeepSeek Harness (DSH), with verbatim evidence and one-by-one approval
Agent-decided approvals for DeepSeek Harness: a workspace-write base permission mode where an independent approval subagent judges every sandbox escalation (risky operations are rejected), with a configurable approval model and a per-session audit trail in the conversation window's 审批 tab.
Codex-style approval autopilot for DeepSeek Harness: ordered glob rules (allow/ask/deny) plus an AI risk judge (low/medium/high) mapped through a risk tolerance, as an approval answerer.
DeepSeek Harness Web UI soundscape: turn-complete celebration (synthesized chime + confetti burst), blocked/approval alerts, error buzz, send clack and optional typing ambience. Zero audio assets, zero core changes, and a ctx.soundscape service other plugins can call.
Real-time attention alerts for DeepSeek Harness: a global banner plus flashing workspace entries when the agent is blocked (approval request / sandbox denial), with an audible chime and OS-level notifications for new blockers.
DSH 任务提醒插件:任务结束/审批/提问/出错时网页 toast + 跨窗口系统弹窗 + 通知中心;设置页内置配置预设,可自定义弹窗位置、样式、主题、宽度与系统级提醒开关
Browser audio alert for human-attention events: approval requests, ask-user questions, and finished turns, with per-trigger sounds configurable through the audio-alert settings namespace
Independent Codex Guardian-style approval reviewer for DSH.
VS Code (from a code-server release) inside DSH: a right-sidebar tab driven by the plugin's own launcher over the in-process VS Code server (lib/launcher.mjs). Since 0.3.0 the bundle also ships an editor bridge (assets/extensions/dshcs-editor-bridge): a read-only channel between the in-tree VS Code extension host and DSH, giving the agent what only the editor knows (unsaved buffers, language-server diagnostics, the active selection) and letting editor gestures drive the session. Since 0.3.22 the ask panel renders that session's new content with DSH's own Markdown renderer (pinned to the deployed DSH UI version); since 0.3.23 it is a full-width dialog with collapsible thinking rows and can answer approval requests in place — the bridge's only non-read-only route (POST /approve, constrained to pending in-process requests and the outcomes allowed-once/rejected; panel-first for 5 minutes, handed back to the official card the moment the panel closes). The tab claims DSH file addresses (dsh-resource://file/**) by file type (setting claimExtensions), so the product's own produced-file chips, delivered-file previews and inline prose mentions open in the workbench. The IDE is a resident surface moved with Element.moveBefore instead of being remounted, so switching sidebar tabs no longer reloads it. Opening the tab switches the right sidebar to fullscreen by default (setting fullscreenOnOpen). Following a workspace switch is lightweight: the workbench re-navigates with the new ?folder= and the IDE process is not restarted (since 0.2.12). Requires a DSH with the right-sidebar services (sidebarRightTabs/sidebarRight, >= 0.1.5-alpha.1); older DSH versions get a single upgrade notice on the settings page and no other UI. Two serving modes: loopback port (default) or same-origin mount on DSH's own webServer (/code-server, protected by ctx.connection.requestRejection). No code-server Node layer, no argon2, no C++ toolchain.
Semantic notification sounds for DeepSeek Harness: complete/approval/question/block/error via BEL or WAV, with Web UI bell toggle | DeepSeek Harness 语义化通知声音:完成/审批/提问/受阻/错误,经 BEL 或 WAV 播放,带 Web 铃铛开关
DeepSeek Harness plugin: panel hotkeys for every interaction panel — Enter presses the confirm (primary) button, Esc presses the cancel button (approval, question/choice, plan review; Claude Code habits).
Require a one-shot user approval before configured tools execute
A middle permission tier for DeepSeek Harness between workspace-write and danger-full-access: auto-approves harmless commands and operations targeting configured trusted areas, beyond the current workspace.
Mi Home (米家) control for DeepSeek Harness agents — list homes/devices, read props, and control devices behind a human approval gate.
Claude Code CLI as an LLM provider for dsh (DeepSeek Harness): live model list, per-session resume, approval relay, images, a memory/rewind/changes panel, and workspaces on remote SSH boxes
Autonomous (auto) mode permission classifier for DeepSeek Harness: a Claude-Code-auto-mode-like classifier over tools/pre-execute and approval/request, a selectable 'auto' permission preset, LLM semantic judge, git checkpointing, agent discipline guidance, and a web control page in Settings → Plugins.