fraud-watch

仓库创建 2026年7月6日最近提交 21 天前SkillHot 收录 20 天前

▸ 精选理由

可处理表格/图片/粘贴数据，适合快速排查可疑交易。

▸ 风险提示

会处理敏感财务与交易数据，注意隐私与上传风险。

这个 Skill 做什么

分析交易记录，识别可能的欺诈与异常消费并给出调查建议。

把你上传或粘贴的交易数据（PDF/CSV/图片/明细行）当证据来分析，自动识别与账户“正常”模式不符的可疑交易并标注常见欺诈模式。适合怀疑被盗刷、要复查账单或排查异常消费时使用。不会直接宣称“这是欺诈”，而是给出逐笔调查建议和后续取证步骤，让你做最终判定。

▸ 展开 SKILL.md 英文原文

Review bank or card transactions and surface possible fraud. Works on whatever transaction data the user provides or the agent can read — an uploaded bank statement (PDF/CSV/image), pasted rows, or a connected banking tool. Establishes what "normal" looks like for the account from the data itself, flags the classic fraud patterns, and walks the user through each suspect charge. Never declares fraud on its own — the user always makes the call.

垂直行业欺诈检测交易分析银行对账通用

Stars

Forks

仓库内 Skill

7 日增星

安装 / 使用

给你的 Agent 一句话（通用）

帮我安装这个 skill：https://raw.githubusercontent.com/kelsi-bizer/bizer-ai-agent-skills/main/skills/fraud-watch/SKILL.md

或 curl 直取 SKILL.md

curl -fsSL "https://raw.githubusercontent.com/kelsi-bizer/bizer-ai-agent-skills/main/skills/fraud-watch/SKILL.md"

SKILL.MD 节选查看完整文件 ↗

# Fraud Watch

You are reviewing bank transactions for signs of fraud, the way a fraud
analyst would. The user may have uploaded a statement, pasted transactions,
or connected an account — it does not matter. Work with whatever transaction
data is in front of you. If you have dates, amounts, and merchant names, you
have enough. Only review data you can actually see; never invent or assume
transactions.

The patterns and thresholds below were calibrated against millions of labeled
fraud transactions. They are rules of thumb to reason with, not a program to
execute — when a charge is borderline, say it's borderline.

## Step 1 — Learn this account's normal

Before hunting anomalies, read the w

via SKILL·HOT · 数据来自 GitHub 公开信息 · 原文版权归作者所有