1337 results
DeepSeek Harness plugin that locks the web chat composer Enter-send. Toggle with a lock button or Ctrl+Shift+K to prevent accidental sends while typing or editing long prompts; unlock to send. IME-safe and per-session.
A portable coding lifecycle for AI agents with deterministic routing across 25 skills, durable state, mutation-aware verification, specialized skill packs, lifecycle hooks, DeepSeek Harness plugin & consumed web UI, and an optional local request proxy.
In-place three-decimal cache-hit percentage plus a widened composer stats line for DSH Web.
A DeepSeek Harness (dsh) plugin: orchestrates complex tasks across multiple models — a 'main brain' decomposes the task and dispatches subtasks to GLM / Kimi / Qwen etc. sub-agents by each model's strengths, with per-model token usage.
生产级中文编剧创作与剧本诊断系统。覆盖概念超短片、短片、电影长片、剧集/连续剧、竖屏短剧/微短剧；强调主角驱动因果、场景价值转折、潜台词、中文自然度、故事状态机、人物知识边界、关系连续性、承诺-兑现、证据式QC、单调修复与长篇记忆快照。用于从零开发、续写、改戏、对白、剧本医生、结构诊断与长篇连续性维护。
dsh plugin: complete Office document tools for the agent (create/read/edit .docx/.xlsx/.pptx through the OfficeCLI binary) plus a document preview in the dsh web sidebar via dsh-better-sidebar.
Global frosted-glass (backdrop blur) theme plugin for the dsh web UI with animated light/dark switching, a scheduled theme timer, a composer glow lighting effect, system notifications and a Monet (Material You) wallpaper color-extraction theme
Selection quoting for the dsh web GUI: a toolbar over selected assistant text that attaches the selection to the composer as a removable quote banner, prepended to the outgoing message at send
Host-level vision bridge for text-only models: analyze_image tool (Ollama local / Xiaomi MiMo cloud / any OpenAI-compatible endpoint) returning structured evidence.
Common Word Box (常用词盒子) — DeepSeek Harness web client plugin: a persistent slide-up word/phrase panel next to the composer input, with global/current-project buckets and one-click insert. Works in every workspace.
Upload arbitrary local files from the DeepSeek Harness Web composer, attach them to prompts, and manage stored uploads in Settings.
提示词增强（DSH 通用插件）：三种模式（基础/标准/专家）读取会话工作区项目上下文并调用 LLM 增强输入框提示词，支持结果对比确认、采用后撤回、失败模型回退链、按模型能力自动适配推理档位（网关不支持时自动去参重试）、待确认问答闭环（逐条回答后带答案重新增强）、组合行 config 可配置、可取消。Host 半提供 /api/prompt-enhance/enhance；浏览器半在 composer 工具行注册模式选择与 ✦ 按钮、对比面板与忙碌状态卡。随 harness 启动自动加载，无需手动重新加载。
DSH 中转站模型思考强度配置与自动识别插件
Auto-inject user-chosen skills (e.g. caveman, ponytail) into DeepSeek Harness sessions: every prompt or once at session start, with a settings page and a composer indicator. 在 DSH 会话中自动注入所选技能（如 caveman、ponytail）：每轮提示或仅在会话开始时，含设置页与输入区指示。
GrayPrint — live reasoning-style fingerprints for the DeepSeek Harness web chat. Compares grayscale-vs-current writing style by internal reasoning paragraphs and independently tracks Let me exploratory openers versus We need planning openers.
Give DeepSeek (text-only) models vision in Claude Code, Codex, and Agent Plugins clients: describe images via any OpenAI-compatible vision endpoint.
输出 token 压缩：让 agent 用紧凑的穴居人语回复（省 ~65% 输出 token），代码/命令/错误保持字节精确；token 下沉点审计（caveman learn：safe fix/offload/habit/load-bearing 分类）。受 JuliusBrussee/caveman（86k★ MIT）启发。
DSH tool-call compatibility: strips sandbox_permissions/justification from tool-call arguments for third-party models (GPT etc.), and lets the user skip possibly-stuck tool calls with an LLM-facing notice.
Tacit learns what you leave unsaid in your prompts — from messy turns and your own corrections, with zero clicks — and tells the agent how to compensate, on every turn, via a system-prompt section you can read and edit.
Complete workspace grouping manager for the DSH web sidebar: 分类 → 项目 → 会话 three-level tree, manual group create/rename/delete, drag-and-drop grouping & ordering with insertion indicator, rule-based auto-classification, tree search. Sidecar YAML + runtime overlay, zero core-storage intrusion.
图片识别插件：自动判断当前模型是否具备视觉能力，有则用当前模型并按插件预设提示词分析，无则调用插件配置的视觉模型。文本模型可直接在对话框粘贴/上传/拖拽图片，发送时自动写入 DSH 附件存储（永久），消息区渲染缩略图，模型自动调用 image_vision / ocr / ground / crop 系列工具识别与精读；模型选择器保持简洁。
A Symphony-compatible multi-provider task orchestrator and dashboard for DeepSeek Harness
Fold mis-rendered thinking content into collapsible blocks, render mermaid inline, show file-edit diff counts, rewrite composer drafts with the current model, and resume interrupted turns from the send button.
Play a short chime when any conversation turn finishes. Built-in tones plus one custom upload, configured in Settings → Plugins.