Back to skill

Security audit

资产负债表日后事项·逐条挂源

Security checks for vulnerabilities and agentic risk

Overview

ClawScan rates this skill suspicious because scanners report behavior and external-service workflows that do not clearly match its stated financial-reporting purpose; the scan was partial, so its evidence is incomplete.

Consider the reported credential routing, external document sharing, and home-directory usage marker before installing. The verdict is suspicious, not a claim that the author is malicious. Scanning was partial: 3 of 4 components were scanned, and one static analysis was degraded, so the evidence is incomplete.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (10)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

所示代码没有读取财报、提取披露原句或生成可溯源事项登记表,而是在执行外部 MCP 服务发现和认证工具枚举。代码也没有展示这些操作与所声明财报分析功能之间的联系,因此不能将其认定为该功能的普通实现细节。需注意,解析出的 routing 仅被打印,未在此代码中赋值给 ROUTING_URL;实际认证请求的目标取决于已有环境变量。

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

代码的可见行为是获取并打印外部 MCP 目录,没有读取财报、抽取日后事项、保留披露原句或生成带锚点的登记表。该目录请求与声明任务没有可见的支持关系,因此构成行为与用途不符;判断并非仅基于网络请求或临时文件写入这些实现细节。

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The described checker records dated usage markers in the user's home directory, but SKILL.md does not disclose that persistent side effect or connect it to document validation. This is a limited privacy and filesystem-boundary issue, not demonstrated remote telemetry or a high-impact compromise.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
91% confidence
Finding

Fetching a public catalog is not itself data exfiltration, but the subsequent instructions attach CUE_API_KEY to a destination obtained from that catalog without specifying destination validation. A compromised catalog or incorrectly supplied ROUTING_URL could therefore direct the bearer credential to an attacker-controlled endpoint.

Content

Scanner excerpt · README.md (reported line 75)May include surrounding context.

发现先行,匿名取 live 目录,拿到 routing 后再连接发 tools/list,以现场返回为准,不硬编码工具名与工具数:

bash
curl -sS https://cuecue.cn/api/mcp-catalog -o /tmp/mcp-catalog.json
python3 -c "import json;d=json.load(open('/tmp/mcp-catalog.json'));print(d.get('routing') or d)"
# 拿到 routing 后,带 token 发 tools/list:
curl -sS -X POST "$ROUTING_URL" -H "Authorization: Bearer $CUE_API_KEY" \

The documented data-channel commands send HTTP requests despite the explicit claim that the skill sends none.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The README provides network-request commands, including an authenticated request, but later states without qualification that the skill makes no requests. Although these are optional workflow instructions rather than automatically executed code, the contradiction can cause users or agents to misunderstand the workflow's network and credential exposure.

Content

No source excerpt is available for this finding.

The workflow directs financial attachments and scans to an external parsing service without warning that their contents will be shared with that provider.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The workflow directs attachments and scans to an external parser without an explicit disclosure or consent step for sharing their contents. Published financial reports reduce the confidentiality risk, but user-supplied attachments or screenshots can contain nonpublic financial, legal, or personal information.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The attachment-parsing workflow lacks a warning that user-provided document content will be sent to an external service.

Medium
Category
Output Handling
Confidence
93% confidence
Finding

The skill instructs the agent to invoke external parsing when relevant text is in an attachment, without requiring approval to disclose the attachment's contents. Many intended inputs are public disclosures, but the accepted screenshots and user-provided documents can also contain confidential financial or legal information.

Content

No source excerpt is available for this finding.

Marking a financial-impact entry as self-calculated bypasses checks documented as prohibiting self-calculation.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

A line containing 〔自算〕 is skipped before financial-amount and adjustment-direction validation, despite the stated prohibition on self-calculated impacts. Because report text controls this marker, an author can label a fabricated amount or reversed treatment as self-calculated and suppress these checks; related exemptions also weaken quotation and classification checks.

Content

No source excerpt is available for this finding.

The checker persistently records usage in the user's home directory without a purpose-related need.

Low
Category
Excessive Agency
Confidence
99% confidence
Finding

mark_first_run creates a persistent, dated skill-usage marker under ~/.cue-usage, including for demo and self-test runs, without needing that state for validation. The marker exposes limited local activity information and silently modifies the user's home directory; the supplied code does not transmit it over the network.

Content

No source excerpt is available for this finding.