AI Hackathon 产品AI hackathon product· 2026
ResetAgent
开发者状态恢复协议Developer state recovery protocol
为疲劳、焦虑、卡住的开发者生成 3 分钟恢复流程和 Agent Handoff Prompt。A three-minute recovery flow and agent handoff prompt for developers who are tired, anxious, or stuck.
案例摘要Case summary
目标用户Audience
高压开发、黑客松、AI 编程中出现疲劳、焦虑或卡住状态的开发者。Developers in high-pressure coding, hackathons, or AI-assisted work who hit fatigue, anxiety, or stuck states.
我的角色Role
个人参赛，负责产品定义、交互流程、双模方案、部署选择和演示取舍；全部代码由 AI 编程工具完成。Solo entrant; owned product definition, interaction flow, dual-mode approach, deployment choices, and demo tradeoffs; all code was produced with AI coding tools.
交付结果Outcome
24 小时内交付可运行网站、项目海报和公开仓库，并保留 API 失败后的本地规则兜底。Shipped a runnable website, project poster, and public repo within 24 hours, with a local rule-based fallback if the API fails.
产品洞察Product insight
疲劳开发者缺的不是更多计划，而是一个低认知负担的停顿机制。产品价值在于让用户先恢复判断力，再决定下一步。Tired developers do not lack more plans — they lack a low-cognitive-load pause mechanism. The product value is letting users restore their judgment first, then decide the next step.
项目简介Overview
ResetAgent 是 24 小时黑客松项目，帮助开发者在高压场景中判断继续、休息还是交给 Agent。它把抽象建议拆成可执行流程，并生成结构化交接 Prompt。ResetAgent is a 24-hour hackathon project that helps developers judge whether to continue, rest, or hand off to an agent in high-pressure situations. It turns abstract advice into an executable flow and generates a structured handoff prompt.
技术栈Tech stack
结果证据Evidence
线上 Demo 已部署到 Vercel。The live demo is deployed on Vercel.
仓库 assets 中保留 desktop、mobile、poster 和 demo QR 等提交材料。The repo assets keep desktop, mobile, poster, and demo QR submission materials.
协议生成引擎支持 LLM 失败后的本地规则兜底，保证 Demo 可用。The generation engine supports a local rule-based fallback on LLM failure, keeping the demo usable.
项目截图Screenshots
关键决策Key decisions
黑客松现场网络和 LLM API 都可能不稳定。Both the network and the LLM API can be unstable at a hackathon.
设计双模生成引擎：优先 LLM，失败时回退到本地启发式规则。Design a dual-mode generation engine: prefer the LLM, fall back to local heuristic rules when it fails.
疲劳用户没有精力读长说明或填复杂表单。A fatigued user has no energy to read long explanations or fill complex forms.
重构为 4 屏渐进式向导，每屏只处理一个判断动作。Refactor into a 4-screen progressive wizard where each screen handles one decision action.
想看完整的代码和实现细节？Want the full code and implementation details?
在 GitHub 上查看View on GitHub