Agent Skills › win4r/MuseAI-Skills › wide_research

wide_research

GitHub

用于协调子代理并行执行大量独立研究任务，将输入拆解后统一返回标准化结果集。适用于批量信息检索、筛查或需要高吞吐量且各子任务互不依赖的场景。

opt/hatch/skills/wide-research/SKILL.md win4r/MuseAI-Skills

Trigger Scenarios

用户请求宽泛研究需要进行大规模并行数据查询涉及多个独立实体的批量信息收集

Install

npx skills add win4r/MuseAI-Skills --skill wide_research -g -y

More Options

Non-standard path

npx skills add https://github.com/win4r/MuseAI-Skills/tree/main/opt/hatch/skills/wide-research -g -y

Use without installing

npx skills use win4r/MuseAI-Skills@wide_research

指定 Agent (Claude Code)

npx skills add win4r/MuseAI-Skills --skill wide_research -a claude-code -g -y

安装 repo 全部 skill

npx skills add win4r/MuseAI-Skills --all -g -y

预览 repo 内 skill

npx skills add win4r/MuseAI-Skills --list

SKILL.md

Frontmatter

{
    "name": "wide_research",
    "metadata": {
        "includeInPrompt": false
    },
    "description": "Use when the user needs broad parallel research across many independent inputs with a shared output schema."
}

Wide Research

Purpose

Coordinate one manager subagent that fans out the same research task across many independent inputs and returns a single normalized result set.

When to Use

The task can be split into many independent subtasks (for example, one company/person/topic per input).
Each subtask should return the same structured fields.
The user asks for "wide research", "parallel research", or high-volume lookup/screening.

Do not use this workflow for single-item tasks or when subtasks depend on each other.

Tooling

Use exactly one manager subagent for the overall operation.

Manager input contract:

operation_brief: one concise sentence.
inputs: one independent item per element.
output_schema: required fields and allowed types.
worker_prompt_template: per-input instructions.
completion_format: exact JSON object shape for the manager's final response.

Manager output contract:

total
success_count
failure_count
results
failures
notes

Operating Rules

Deduplicate and normalize the input list before spawning the manager.
Keep the output schema minimal and explicit; avoid optional or free-form fields unless the user asked for them.
Use one manager coordinator for one user goal; do not fan out multiple sibling root-level subagents.
Instruct the manager to assign one worker per input and keep each worker scoped to its own item.
If completeness matters and some inputs fail, retry only the failed inputs, once, when feasible.
Reply to the user immediately that wide research has started; do not block on completion.
In the final user-facing output, always report coverage (success_count/total) and unresolved gaps.