203 results
DeepSeek Harness 自动技能策展插件：后台评审子代理定期复盘对话，自动创建/更新 SKILL.md，让智能体在真实使用中持续自我进化，零侵入、不改 dsh 源码。Automatic skill curation for DeepSeek Harness — a background subagent reviews conversations periodically and creates/updates SKILL.md files; zero-intrusion bundle plugin, no dsh source changes.
DSH skill bundle: competition math (IMO/Putnam/USAMO/AIME) solved with a pure-reasoning pass, an adversarial verifier in a fresh subagent context attacking concrete failure modes, and calibrated confidence output (high / medium / honest "no confident solution"); optional LaTeX→PDF rendering.
DeepSeek official-home style theme: fluid particle background, glass sidebar, glass composer, frosted header, running-subagent progress panel, and DeepSeek brand link for the dsh web GUI
Fixed five-member dev team for DeepSeek Harness driven by the spec-superflow state machine: captain + spec-writer/architect/engineer/reviewer/docs, phase-tagged tasks with dependencies, auto phase advancement with a human build gate, and a workflow stepper in the web GUI
AI-powered command approval plugin for DeepSeek Harness — intercepts tool calls and routes them through a read-only review subagent
DSH plugin: render every subagent as a clickable card in the main conversation that opens the child as an ordinary session — OpenCode-style subagent visibility, without re-rendering the shell's session view.
Cursor-as-subagent provider for DeepSeek Harness: one-shot local runs via @cursor/sdk, summary-first result presentation, unattended Profile Bundle.
DSH plugin: agent-schedulable recurring timer tasks with veto. When a task fires, the owning session (or its blocking subagent) is prompted and may veto (timer_veto tool or marker file); the command runs only when nobody vetoes.
让主会话模型从已配置模型里自主选择子代理模型
AgentTeams for DeepSeek Harness: multi-agent team collaboration (captain, members, tasks with dependencies, messaging) driven by natural language, with a tree monitor in the web GUI
对话流实时任务进度：后台任务（pwsh/bash/subagent 等 run_in_background）显示状态、耗时、进度条、预计剩余时间与停止按钮；非消费式 jobs.get 跟踪，模型 job_output 不受影响。
Per-session subagent model picker for DeepSeek Harness with immutable child bindings
并行子代理集群：母代理把目标拆成切片，并发驱动多个独立子代理写代码，实时可视化每个子代理在做什么，并在设置里配置并发数与各切片的模型。
Run every DeepSeek Harness task directly in the FlowText Agent UI
Processes tab for DeepSeek Harness: a live per-process console (main agent and every subagent) with raw tool requests, responses and timing, installed as one profile bundle.
并行多角色代码评审小队：/review 一条命令派出安全/正确性/测试/风格四名只读评审员子代理（每人可单独指定模型与思考强度），汇总为按严重度分组的结构化报告。Parallel multi-role code review squad: /review dispatches security/correctness/tests/style reviewer subagents (each on its own model) and aggregates one severity-grouped report.
Claude Code as a real, streaming dsh subagent: session-backed child, free model/effort choice per call, live model list from the CLI
DSH 对抗多智能体三层失效的防御插件：读取台账去重度量(read-ledger)、免疫压缩规则区+完成机械门禁(immutable-core+completion-gate)、低成本审查lane(role-router)。零外部依赖，注入即用。
Run Claude Code headless as a subagent from the DeepSeek Harness, streaming a live digest of its steps into the session view while it works.
Group work for DeepSeek Harness: turn one session into a room with a chair and several member agents that message each other and share a task list.
One-shot Codex subagent provider over the official app-server protocol
Limit concurrent DeepSeek Harness subagent turns and delegation runs
Ships a skill that tells the agent when to delegate work to subagents instead of doing it inline, how to fan several out in parallel, and how to check what they returned.
dsh web Settings section (Custom models): subagent delegation allow-list, plus per-model image-input / reasoning-effort mapping.