Skills for Real Engineers. Straight from my .agents directory.
research-deep
Agent skill fromWeizhena/Deep-Research-skills
读取调研outline,为每个item启动独立agent进行深度调研。禁用task output。
How to install research-deep
npx skills add Weizhena/Deep-Research-skills# or add only this skill to Claude Code: copy the folder
skills/research-zh/research-deep/
# into
~/.claude/skills/research-deep/ (all projects)
.claude/skills/research-deep/ (one project)
1# Research Deep - 深度调研23## 触发方式4`/research-deep`56## 执行流程78### Step 1: 自动定位Outline9在当前工作目录查找 `*/outline.yaml` 文件,读取items列表、execution配置(含items_per_agent)。1011### Step 2: 断点续传检查12- 检查output_dir下已完成的JSON文件13- 跳过已完成的items1415### Step 3: 分批执行16- 按batch_size分批(完成一批需要得到用户同意才可进行下一批)17- 每个agent负责items_per_agent个项目18- 启动web-search-agent(后台并行,禁用task output)1920**参数获取**:21- `{topic}`: outline.yaml中的topic字段22- `{item_name}`: item的name字段23- `{item_related_info}`: item的完整yaml内容(name + category + description等)24- `{output_dir}`: outline.yaml中execution.output_dir(默认./results)25- `{fields_path}`: {topic}/fields.yaml的绝对路径26- `{output_path}`: {output_dir}/{item_name_slug}.json的绝对路径(slugify处理item_name:空格替换为_,移除特殊字符)2728**硬约束**:以下prompt必须严格复述,仅替换{xxx}中的变量,禁止改写结构或措辞。2930**Prompt模板**:31```python32prompt = f"""## 任务33调研 {item_related_info},输出结构化JSON到 {output_path}3435## 字段定义36读取 {fields_path} 获取所有字段定义3738## 输出要求391. 按fields.yaml定义的字段输出JSON402. 不确定的字段值标注[不确定]413. JSON末尾添加uncertain数组,列出所有不确定的字段名424. 所有字段值必须使用中文输出(调研过程可用英文,但最终JSON值为中文)4344## 输出路径45{output_path}4647## 验证48完成JSON输出后,运行验证脚本确保字段完整覆盖:49python ~/.claude/skills/research/validate_json.py -f {fields_path} -j {output_path}50验证通过后才算完成任务。51"""52```5354**One-shot示例**(假设调研GitHub Copilot):55```56## 任务57调研 name: GitHub Copilot58category: 国际产品59description: Microsoft/GitHub开发,首个主流AI编程助手,市场份额约40%,输出结构化JSON到 {project_dir}/results/GitHub_Copilot.json6061## 字段定义62读取 {project_dir}/fields.yaml 获取所有字段定义6364## 输出要求651. 按fields.yaml定义的字段输出JSON662. 不确定的字段值标注[不确定]673. JSON末尾添加uncertain数组,列出所有不确定的字段名68…
Excerpt shown under the repository's MIT license.
More skills in Deep-Research-skills
Similar skill repositories
An agentic skills framework & software development methodology that works.
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.
Public repository for Agent Skills
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
