Files
Szpitale-graph/skills/research/deep/SKILL.md
T
jilinchenandClaude Opus 4.5 302f2733d3 refactor: separate outline.yaml and fields.yaml
- Output to {topic}/ directory structure
- fields.yaml now independent file for field definitions
- outline.yaml contains items + execution config
- Deep research agents read fields.yaml independently (hard constraint)
- Update README with new directory structure

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2025-12-30 10:42:58 +08:00

44 lines
1.3 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
---
description: 读取调研outline,为每个item启动独立agent进行深度调研。禁用task output。
allowed-tools: Bash, Read, Write, Glob, WebSearch, Task
---
# Research Deep - 深度调研
## 触发方式
`/research-deep`
## 执行流程
### Step 1: 自动定位Outline
在当前工作目录查找 `*/outline.yaml` 文件,读取items列表、execution配置。同时读取同目录下的 `fields.yaml` 获取字段定义。
### Step 2: 断点续传检查
- 检查output_dir下已完成的JSON文件
- 跳过已完成的items
### Step 3: 分批执行
- 按batch_size分批(默认5个)
- 每个item启动1个web-search-agent(后台并行,禁用task output
- **硬约束**:仅传入item相关信息和输出路径
- **硬约束**:agent必须自行读取 `{topic}/fields.yaml` 获取字段定义
- **硬约束**:禁止在prompt中直接嵌入字段定义
- 输出结构化JSON到output_dir
### Step 4: 等待与监控
- 等待当前批次完成
- 启动下一批
- 显示进度
### Step 5: 汇总报告
全部完成后输出:
- 完成数量
- 失败/不确定标记的items
- 输出目录
## Agent配置
- 每批并行: 5个agent(可在outline中配置)
- 后台执行: 是
- Task Output: 禁用(agent完成时有明确输出文件)
- 断点续传: 是