Token导航 LogoToken导航

Agent Skill · 研究检索

tax-fraud-detection(税务欺诈检测)

tax-fraud-detection 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

tax-fraud-detection 是研究检索类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为需要联网;许可证为 MIT,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 67,168。

总安装 250周安装 10GitHub Stars 4下载 81核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:tax-fraud-detection(税务欺诈检测)
来源仓库:https://github.com/cleanexpo/ato
仓库路径:skills/tax-fraud-detection
安装命令:
npx skills add https://github.com/cleanexpo/ato --skill tax-fraud-detection
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/cleanexpo/ato --skill tax-fraud-detection

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

Tax Fraud Detection 欺诈识别模块,利用模式识别发现可疑申报行为。适用于税务局端风险筛查或企业内部审计辅助工具。基于 cleanexpo/ato 架构开发,可能采用机器学习算法实现。存在误判可能,最终结论需经人工复核确认。

tax-fraud-detection 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

来源仓库内路径:skills/tax-fraud-detection

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/cleanexpo/ato --skill tax-fraud-detection;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要需要联网,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

tax-fraud-detection 怎么安装?

在终端执行 npx skills add https://github.com/cleanexpo/ato --skill tax-fraud-detection。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

tax-fraud-detection 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

tax-fraud-detection 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

tax-fraud-detection 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

tax-fraud-detection 用的人多吗?

公开数据是累计安装 250 次,近一周安装 10 次,下载 81 次,GitHub 4 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

tax-fraud-detection 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。
  • 第三方扫描已通过,常见的高风险行为没有命中。

先想清楚的地方

  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
tax-fraud-detection

税务欺诈检测 · 本页

2504需要联网MIT通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Tax Fraud Detection Skill

Proactive detection and prevention capability ensuring no recommendation produced by the Australian Tax Optimizer platform constitutes illegal tax avoidance or would trigger Part IVA anti-avoidance provisions.

When to Use

Activate this skill when:

  • Reviewing tax recommendations for legal defensibility
  • Analysing transaction patterns for circular or artificial arrangements
  • Assessing R&D claims for legitimacy (not routine activities disguised as R&D)
  • Evaluating Division 7A arrangements for sham characteristics
  • Checking loss carry-forward for artificial loss generation
  • Assessing trust distribution arrangements under Section 100A
  • Any recommendation involves > $50,000 in claimed benefits

Part IVA Anti-Avoidance Framework

Overview (ITAA 1936, Part IVA)

Part IVA is the general anti-avoidance provision. The Commissioner can cancel a tax benefit if:

  1. A scheme exists (s 177A) - Very broadly defined; almost any arrangement qualifies
  2. A tax benefit was obtained (s 177C/CB) - Amount not included in assessable income, or deduction claimed
  3. Dominant purpose was tax benefit (s 177D) - Assessed objectively, considering 8 factors

The 8 Factors (s 177D(2))

FactorDescriptionRed Flag
(a)Manner of schemeUnusual or contrived steps
(b)Form and substanceForm doesn't match economic substance
(c)Time of entryArrangement timed around FY boundaries
(d)Result achievedResult only makes sense if tax benefit obtained
(e)Change in financial positionParties' positions unchanged except for tax
(f)Change in financial position (others)Related parties benefit disproportionately
(g)Tax consequencesScheme designed to exploit specific provisions
(h)Connection between partiesRelated party involvement

Red Flags

R&D Tax Incentive Fraud Indicators

IndicatorRisk LevelAction
Routine software development claimed as R&DHIGHVerify technical uncertainty existed
IT operations/maintenance claimed as R&DHIGHVerify systematic investigation occurred
Cloud hosting/infrastructure claimed as R&DMEDIUMVerify experimental purpose
Consulting fees claimed as R&DMEDIUMVerify consultant performed R&D activities
No technical uncertainty documentedCRITICALCannot claim - Division 355 requires uncertainty
Activities completed before registrationHIGHMay indicate retrospective claim

Division 7A Fraud Indicators

IndicatorRisk LevelAction
Circular loan arrangementsCRITICALLoan → dividend → repayment → new loan
Loans at zero interestHIGHDeemed dividend unless compliant agreement
No written agreementHIGHMust exist before lodgment date
Loans to associates/related trustsMEDIUMCheck UPE and sub-trust arrangements
Artificially low repaymentsMEDIUMBelow minimum yearly repayment
Backdated loan agreementsCRITICALFraud indicator

Deduction Fraud Indicators

IndicatorRisk LevelAction
Private expenses claimed as businessCRITICALNo nexus to income production
Capital expenditure claimed as revenueHIGHMust be capitalised and depreciated
Duplicate deductions across entitiesCRITICALSame expense claimed by multiple entities
Inflated amounts vs market valueHIGHTransfer pricing / Part IVA risk
Related party payments above marketHIGHNon-arm's length dealing

Loss Carry-Forward Fraud Indicators

IndicatorRisk LevelAction
Artificial losses via uncommercial dealsCRITICALPart IVA cancellation of loss
Loss trafficking (company acquisition for losses)CRITICALCOT/SBT specifically prevents this
Circular transactions generating lossesCRITICALSham arrangement
Losses from non-commercial activitiesHIGHDivision 35 restrictions apply
Losses from tax shelter arrangementsHIGHDivision 245 debt forgiveness rules

Process

Step 1: Scan

Examine the recommendation or transaction pattern:

  • Extract all claimed tax benefits (amounts, provisions)
  • Identify parties involved (related? associated? independent?)
  • Map the arrangement steps chronologically
  • Note any unusual timing or structure

Step 2: Classify

Categorise the arrangement:

CategoryDescriptionRisk
Ordinary commercialNormal business transactionLOW
Tax-effectiveStructured for tax efficiency, genuine commercial purposeLOW-MEDIUM
AggressiveTax benefit is significant motivation, some commercial substanceMEDIUM-HIGH
ArtificialNo genuine commercial purpose, designed for tax benefitHIGH-CRITICAL
ShamParties never intended to give effect to arrangementCRITICAL

Step 3: Assess

Apply the Part IVA eight-factor test:

  • Score each factor 0-10 (0 = no concern, 10 = maximum concern)
  • Overall score: Sum / 80 × 100 = percentage risk
  • Threshold: > 40% = flag for professional review
  • Threshold: > 70% = do not recommend

Step 4: Report

Generate findings:

  • Specific risk identified
  • Part IVA factor(s) triggered
  • Recommended action (proceed / modify / abandon)
  • Legislative basis for concern
  • Professional review requirement

Output Template

<fraud_assessment>
  <arrangement>Description of arrangement being assessed</arrangement>
  <date>YYYY-MM-DD</date>

  <part_iva_analysis>
    <scheme_exists>true|false</scheme_exists>
    <tax_benefit_amount>$X</tax_benefit_amount>

    <factors>
      <factor id="a" name="Manner of scheme" score="0-10">
        <assessment>Analysis</assessment>
      </factor>
      <!-- Factors b through h -->
    </factors>

    <overall_score>Percentage</overall_score>
    <risk_level>low|medium|high|critical</risk_level>
  </part_iva_analysis>

  <red_flags>
    <flag severity="critical|high|medium|low">
      <description>Specific concern</description>
      <legislation>Relevant provision</legislation>
    </flag>
  </red_flags>

  <recommendation>
    <action>proceed|modify|abandon</action>
    <conditions>What must be true to proceed safely</conditions>
    <professional_review_required>true|false</professional_review_required>
  </recommendation>
</fraud_assessment>

Key Case Law

CasePrincipleApplication
FCT v Spotless Services (1996) 186 CLR 404Dominant purpose test is objectiveDon't rely on stated commercial reasons
FCT v Hart [2004] HCA 26Part IVA applies to individual stepsEach step of arrangement assessed
FCT v Futuris Corporation [2008] HCA 32Commissioner's discretion is broadPart IVA applies broadly
Harding v FCT [2019] FCAFC 29Interposed entity arrangementsApplicable to trust/company structures

Compliance Standards

  1. Never recommend arrangements primarily motivated by tax benefit - Commercial substance required
  2. Always flag related party transactions - Non-arm's length risk
  3. Verify R&D activities are genuine - Not routine operations relabelled
  4. Ensure losses are commercially generated - Not artificial or circular
  5. Document commercial purpose - For every recommendation, state the non-tax reason
  6. Conservative approach - When in doubt, recommend professional review

平台分布

Claude Code

28.33%
按下载量换算23

Cursor

22.13%
按下载量换算18

windsurf

17.62%
按下载量换算14

OpenCode

12.39%
按下载量换算10

Codex

6.88%
按下载量换算6

Antigravity

3.37%
按下载量换算3

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。

继续浏览同类 Skills