DeepSeek Harness プラグイン — 868 ページ

把左侧会话列表中正在等待审批（待审核）的那一行用红色呼吸动画高亮。

待执行（Pending Session）：把带草稿的新会话留在侧边栏，点进继续编辑，发送即转正为普通对话。草稿文本与附件本地持久化，重启不丢。

Pentesting agent plugin for DeepSeek Harness

Per-model temperature for hand-declared llm-pi-ai providers: a settings page plus an agent/request listener that puts the configured temperature on the wire.

每轮对话独立：每轮开始前把本轮之前的历史从模型可见上下文中折掉，并按会话注入一段长期有效的设定。

百分号编码

百分比计算

百分位数

Webhook notifications for DeepSeek Harness sessions: POST when an approval is requested or a turn ends (ntfy / Bark / generic JSON).

DSH 性能剖析插件：UI 卡顿监测（消息延迟采样）+ procdump/DumpStack 分析 + 托管堆统计

Draggable transparent HUD for DeepSeek Harness showing CPU, GPU and memory usage plus CPU/GPU temperatures.

纯客户端性能 HUD:帧与最差帧、Longtask(四条腿归因)、每插件回调速率与同步耗时、DOM 增长与写入者、JS 堆、宿主 event-loop 与链路 RTT,带隔离实验与一键导出 JSON;守夜层默认常开,用帧级环形缓冲与 LoAF 自动留存最差几次卡顿并导出事故包。 Client-side performance HUD for DeepSeek Harness: FPS and worst frame, long tasks with four-leg attribution (fiber / subtree / DOM write / scheduler), per-plugin callback rate and sync time, DOM growth and writers, JS heap, host event loop and link RTT, plus an isolation experiment and one-click JSON export. An always-on night-watch recorder (per-frame ring buffer + LoAF) keeps the worst stalls and exports an incident pack.

A DeepSeek Harness plugin that shows which plugins are using the host's CPU, memory and disk, attributed by sampling inside the single host process. | DeepSeek Harness 插件资源开销面板：看清哪个插件在占用 CPU、内存与磁盘。

性能优化：测量先行、瓶颈定位、复杂度分析、缓存/懒加载、资源释放。受 addyosmani/agent-skills（88k★ MIT）启发。

DSH 运行监控：侧栏实时读数；整机占比、CPU/内存/界面响应三轨联动、进程趋势与工具进程启停。

Performance diagnosis, repeated plugin isolation campaigns, safe recovery, and measured verification for DeepSeek Harness

A 6-level Codex-style performance slider for the DeepSeek Harness web surface: each level maps a model + reasoning effort and switches through 31 local cutout background frames served by the plugin host.

Free-interval session reports for DeepSeek Harness: AI-narrated daily/weekly/monthly digests over any date range, with every-N-days scheduled reminders that generate the report and pop a system notification (macOS Notification Center / Linux notify-send).

DSH plugin: keep text-only models (deepseek-v4-flash / deepseek-v4-pro) as the session default and automatically route image-bearing requests to a configured vision model; also ships the generic file-attachment feature (drag/paste any file as a durable attachment the agent reads by path) as core patches (apply via scripts/patch-core.mjs).

Permission matrix for DeepSeek Harness: 3 sandbox modes x 4 approval strategies = 9 selectable presets, with a global default, a separate LLM-robot default, three-tier risk policies, audit log and git checkpoint.

Claude Code-style declarative permission rules for DeepSeek Harness: ordered allow/deny/ask rules with tool-name, argument (glob/regex), and workspace-path matching on the tools/pre-execute waterfall, session-log audit, and HMR reload.

Claude Code-style declarative permission rules for DeepSeek Harness: ordered allow/deny/ask rules with tool-name, argument (glob/regex), and workspace-path matching on the tools/pre-execute waterfall, session-log audit, and HMR reload.

Claude Code-style declarative permission rules for DeepSeek Harness: ordered allow/deny/ask rules with tool-name, argument (glob/regex), and workspace-path matching on the tools/pre-execute waterfall, session-log audit, and HMR reload.

Claude Code-style declarative permission rules for DeepSeek Harness: ordered allow/deny/ask rules with tool-name, argument (glob/regex), and workspace-path matching on the tools/pre-execute waterfall, session-log audit, and HMR reload.