Menu
科研idea规划与跨对话追踪skill。将科研需求转化为 Milestone-Keypoint 两级任务体系，以 .plan 文件结构支持跨对话状态追踪。触发场景：用户需要规划实施路径、追踪多阶段进度、跨对话继续科研项目、调整现有计划。
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | idea-plan |
| description | 科研idea规划与跨对话追踪skill。将科研需求转化为 Milestone-Keypoint 两级任务体系，以 .plan 文件结构支持跨对话状态追踪。触发场景：用户需要规划实施路径、追踪多阶段进度、跨对话继续科研项目、调整现有计划。 |
面向科研人员的分布式任务规划系统，对齐论文写作的思维模式。
<idea-name>.plan/
├── INDEX.md # 任务总索引（状态自动同步自 state.md）
├── overview/
│ ├── abstract.md # What: 摘要、动机、研究问题、主要方法
│ └── architecture.md # How: 全局架构、主要管线、模块分解
├── tasks/ # Milestone → Keypoint 层级
│ ├── 1/
│ │ ├── 1.md # Milestone 总述
│ │ ├── state.md # 状态 + 跨对话摘要
│ │ ├── 1.1/
│ │ │ ├── 1.1.md # Keypoint 细述
│ │ │ └── state.md
│ │ └── ...
│ └── ...
└── log/ # 可选
├── chat/ # 用户显式保存的AI对话
└── record/ # Agent汇总的问题/发现
| 层级 | 定位 | 必备字段 |
|---|---|---|
| Milestone | 阶段性里程碑，对应Paper章节 | 说明 |
| Keypoint | 原子任务，可并行执行 | 目标、技术要点、产物 |
Agent可按需添加少量辅助字段（如"前置依赖"、"验收标准"），保持精炼即可。
状态：⬜️未开始 → ⏳进行中 → ✅已完成 → 🔍人工审核
---
state: ⬜️ # ⬜️ | ⏳ | ✅ | 🔍
---
# 摘要
<!-- 跨对话时供Agent快速了解当前进展 -->
# 问题与阻塞
<!-- 实施中遇到的问题，便于下次对话继续 -->
| 阶段 | 触发条件 | 指南 |
|---|---|---|
| 创建 | 已明确需求，无.plan | stage/create.md |
| 读取 | 存在.plan，需恢复上下文 | stage/read.md |
| 更新 | 完成任务，需同步状态 | stage/update.md |
| 迭代 | 需调整idea或技术路线 | stage/iterate.md |
配套 idea-plan-mcp 提供以下工具：
| 工具 | 功能 |
|---|---|
plan_create | 创建.plan骨架（默认1×1最小结构） |
plan_list | 任务树概览 |
plan_read | 分级读取（scope: overview/index/milestone:X/keypoint:X.Y/current） |
plan_update_state | 更新状态+摘要，自动同步INDEX.md |
plan_update_task | 增删改任务（add/delete/rename Milestone或Keypoint） |
plan_sync_index | 手动同步state.md→INDEX.md |
AshAi extension guidelines for integrating AI capabilities with Ash Framework. Use when implementing vectorization/embeddings, exposing Ash actions as LLM tools, creating prompt-backed actions, or setting up MCP servers. Covers semantic search, LangChain integration, and structured outputs.
This skill should be used when solving hard questions, complex architectural problems, or debugging issues that benefit from GPT-5 Pro or GPT-5.1 thinking models with large file context. Use when standard Claude analysis needs deeper reasoning or extended context windows.
Personal assistant for daily routines, task management, and productivity
Create, validate, and convert skills for the agent ecosystem. Enforces standardized structure for consistency. Enables self-evolution by creating new skills on demand, converting MCP servers and codebases to skills.
Rules for working with AshGraphql
AshJsonApi guidelines for exposing Ash resources as JSON:API compliant REST endpoints. Use when adding JSON:API extensions to domains/resources, configuring routes, or implementing filtering, sorting, pagination, and includes. Supports full JSON:API specification.