agent-preset

kixparadigm — AI self-orchestrated minimal paradigm (resident cognition layer) + kixpower multi-agent orchestration · one-command import into DeepSeek Harness (npm i -g) / AI 自编排最小范式（认知层常驻）× kixpower 多智能体编排 · npm 一键导入 DeepSeek Harness

DeepSeek 深度调校的沉浸式角色扮演 Agent 预设（dsh）——零工具纯对话、酒馆式演出格式、文件记忆库

DeepSeek Harness preset: first round uses the official Minimal tools, then unlocks full Standard. 复现 Project2 评测中更高分的 We/Let's 思维链（Minimal 99/96 vs Standard 91/92），改善 Windows 上 DeepSeek V4 Pro 的 Let me 现象。

Two-phase DeepSeek Harness preset: Minimal-aligned bootstrap (bash+read), then full Standard tools after the first tool call or reply

DSH (DeepSeek Harness) 专家模式 agent preset — 统筹团长 + 10 位领域专家子代理 Expert-mode preset for DeepSeek Harness

Minimal Windows agent preset + Git Bash tool for DeepSeek Harness: gitbash & pwsh & str_replace_editor, drop-in replacement for the official minimal preset on win32 / DeepSeek Harness 极简模式（Windows）Agent 预设 + Git Bash 工具：Git Bash + PowerShell + str_replace_editor 三工具，官方 minimal 预设的 win32 平替

Anchored-standard agent preset for DeepSeek Harness: Minimal-aligned first request, low-injection post-promotion, indirect AGENTS.md loading — anchored intelligence without the IQ drop. （DeepSeek Harness 的锚定标准预设：首轮 Minimal 对齐、晋升后低注入、AGENTS.md 间接按需加载——锚定智力不降智。）

dsh 插件：/gamemode 1 = 一键切换到内置创造模式（cordis）预设。dsh plugin: /gamemode 1 = switch to the built-in Creative Mode (cordis) agent preset.

DSH 插件：极简模式（网络）preset——极简双工具编码 Agent + 网络搜索 + AGENTS.md 加载 | Minimal-mode agent preset with web search and AGENTS.md loading

DSH multimodal plugin: drag-image auto-describe with a configurable OpenAI-compatible vision model (host patch + agent preset + optional adapter). MIT.

DeepSeek Harness agent preset: fixes DeepSeek V4 Pro degradation (降智) via RL first-request anchoring — bash+str_replace_editor bootstrap, resident plugin tools, per-turn format guard

Enhanced auto-compaction defaults for DeepSeek Harness agent presets

Vibe Mathematics —— 多代理数学问题求解与形式化验证框架

A DeepSeek Harness (DSH) agent preset that teaches while coding — concrete scenario-grounded explanations, Socratic guidance, and TODO(你) practice blanks, modeled on Claude Code's Learning output style. 学习模式

Professional contract-drafting agent mode for DeepSeek Harness: 11-stage lawyer workflow with 5-way parallel AI review, decision gate, and domain packs (general contract / employment / equity investment)

DeepSeek Harness 的「总监模式」Agent 预设：双层委派架构。主 Agent（总监）运行在高价值模型上，只做判断、计划、任务分解与验收，不写代码；所有子代理固定运行在低成本模型（opencode-go / deepseek-v4-flash）上执行具体任务。maxDepth:1 严格限制两层委派，toolFilter 从执行层移除委派与询问工具，结构上杜绝子代理再开子代理。附 8 项验收清单的 ducheck 验证任务。

First-principles learning assistant preset for DeepSeek Harness: knowledge components with decaying mastery, an analogy lifecycle, and a prediction ledger.

Windows-adapted DSH coding agent preset with persistent PowerShell 7 (pwsh) and str_replace_editor, mirroring the official minimal preset.

Product Team Mode - a DeepSeek Harness agent preset: user-led conversation with a virtual product team (PM to Engineer to QA to Release) walking you from idea to shipped product