AGENT SKILL
cross-border-spike-detector
暂无简介
- License 许可
- MIT
- 最近更新
- 2026-07-21
- Stars
- ★ 835
- Forks
- 83
02 资源文件 3
- 脚本index.mjs4.1 KB
- 参考文档skill.manifest.json2.2 KB
- 参考文档skill.ui.json1.1 KB
03 SKILL.md 正文
Cross-border Spike Detector
Cluster product spike notes from sales, viral creators, paid ads, search trends, and category lifts into a weekly spike review.
When to use
- The cross-border
weekly-hot-product-reviewworkflow is enabled (this skill is the first step in that workflow). - The user pastes weekly spike notes and wants a focused screening of which spikes are worth scouting.
Inputs
spikeSignals(string, required): one observation per line. Mixed Chinese/English notes are supported.
Outputs
spikeReview: a multi-line cluster summary of detected spikes; this output is consumed bycross-border-product-scoutin the chained workflow.summary: one-line headline.nextStep: recommended next operator action; defaults to a risk prompt when IP/compliance language appears.details: structured cluster + risk-signal data for UI rendering.
Behavior
- Deterministic keyword bucketing across sales spike, viral signal, paid ad lift, search trend, and category-wide lift.
- Surfaces explicit IP/compliance risk flags before recommending any sampling/procurement follow-up.
- No external network calls, no filesystem writes, no provider dependency.
- Returns a soft empty bundle when notes are missing instead of inventing findings.