| name | research |
| description | Content topic research and competitor analysis. Activate when user asks to find topics, research competitors, analyze trending content, or generate content ideas for Chinese social media (Xiaohongshu, Douyin, WeChat).
|
Research
Executor skill. Searches trending topics, analyzes competitor content, and saves topics locally.
Research Mode
AutoCrew research is now browser-first:
- Prefer the user's own logged-in browser session through the host runtime's CDP / browser tools.
- Use API providers like TikHub only as fallback.
- If neither is available, degrade to manual/web search based research.
Do not assume TikHub is the default data source anymore.
Inputs
| Parameter | Source | Required | Description |
|---|
| topic_count | Message | No | Number of topics to generate (default: 3) |
| direction | Message | No | Specific topic direction or theme |
| user_requirements | Message | No | Additional constraints |
Steps
-
Read MEMORY.md (if exists at ~/.autocrew/MEMORY.md). Extract: industry, platforms, audience, content style, competitor accounts (Competitor Accounts section).
- IF MEMORY.md is empty or industry is missing THEN ask user for their niche/industry before proceeding.
-
Check for XHS competitor accounts in MEMORY.md.
- IF no XHS competitors THEN go to Step 3C (degraded mode).
- IF XHS competitors exist THEN go to Step 3A (normal mode).
3A. Normal mode / browser-first
Use host-provided browser/CDP capabilities first:
- open the user's logged-in platform session
- inspect competitor recent notes/videos
- inspect platform hot lists
- inspect creator center or search result pages when relevant
Only if browser access is blocked should you drop to Step 3B.
3B. Fallback mode / API provider
If browser-first is unavailable, use TikHub or other structured providers when configured:
- fetch competitor recent posts
- fetch platform content detail when available
- keep a note that results came from fallback API mode
3C. Free mode — 纯公开搜索 + 爆款评分引擎:
使用 autocrew_research tool 的 free mode:
{
"action": "discover",
"keyword": "AI 编程",
"platform": "xiaohongshu",
"mode": "free",
"search_results": []
}
Free 引擎流程:
- 调用方先用
web_search 搜索 3-5 条查询(引擎会生成推荐查询)
- 将搜索结果传入
search_results 参数
- 引擎自动评分(标题吸引力 + 话题热度 + 用户匹配度)
- 风格校准过滤(排除触碰 styleBoundaries.never 的选题)
- 返回排序后的候选选题
推荐搜索查询模板(由 free-engine 的 buildSearchQueries 生成):
{industry} {keyword} 内容选题
{industry} 热门话题 {当前月份}
{keyword} 爆款内容 分析
{keyword} {平台名} 爆款
{keyword} 最新趋势 {年份}
3D. Manual degraded mode — 无任何数据源:
Use web_search only:
- Search:
{industry} 内容选题 小红书
- Search:
{industry} 热门话题 {current_month}
- Search:
{industry} 爆款内容 分析
-
Generate topics. For each topic:
- Title: ≤20 characters, specific angle (not generic)
- Description: WHY this topic has potential + cite the data source
- Tags: 3-5 relevant tags
- Source: where the idea came from (competitor name, trend, hot list)
Prefer using the autocrew_research tool in browser-first mode when available:
{
"action": "discover",
"keyword": "AI 编程",
"platform": "xiaohongshu",
"topic_count": 5
}
-
Quality gate — before saving, each topic must pass:
-
Save topics using autocrew_topic tool:
{ "action": "create", "title": "...", "description": "...", "tags": [...], "source": "..." }
Save each topic individually. Count successes.
-
Output summary:
调研完成,共保存 {saved_count} 个选题:
1. {title_1}
2. {title_2}
...
下一步可以选一个开始写,或者让我继续调研其他方向。
Error Handling
| Failure | Action |
|---|
| Browser session unavailable | Fall back to API provider mode, then degraded mode |
| TikHub unavailable | Continue with browser-first or degraded mode |
| All searches fail | Use general knowledge to generate topics. Note in description that no live data was available. |
| Topic save fails | Log error, continue saving remaining topics. |
Changelog
- 2026-03-31: v1 — Adapted from Qingmo research.md. Removed backend API dependency, uses autocrew_topic tool. TikHub optional, web_search as primary data source.
- 2026-03-31: v2 — Strategy shifted to browser-first research using user-managed login sessions. TikHub moved to fallback provider role.
- 2026-04-01: v3 — Added Free mode (Step 3C) using free-engine.ts. Auto mode now falls back to free engine before manual degradation. New
search_results parameter for pre-fetched web search data.