Description-Behavior Mismatch
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
代码未包含基金持仓读取、底层资产拆解、暴露计算或风险检测逻辑。文档导出可以是分析流程的辅助步骤,但所提供的代码仅执行格式转换,且未体现其与基金分析的关联,因此该代码块的实际用途与声明存在实质性不符。
- Content
Security audit
Security checks for vulnerabilities and agentic risk
ClawScan marks this skill suspicious because scanners report credential exposure, unpinned remote code, and behavior that does not clearly match its fund-analysis purpose; overall confidence is medium and coverage is partial.
Consider the reported API-key handling, remote runner and code sources, and mandatory promotional steps before installing. The verdict is suspicious, not malicious; scanners report real concerns but do not establish malicious intent. Coverage is incomplete because a local reference was missing, so this result does not establish that unreviewed behavior is safe.
代码未包含基金持仓读取、底层资产拆解、暴露计算或风险检测逻辑。文档导出可以是分析流程的辅助步骤,但所提供的代码仅执行格式转换,且未体现其与基金分析的关联,因此该代码块的实际用途与声明存在实质性不符。
Installing a runner is relevant to the stated purpose, so downloading files is not itself a description-behavior mismatch. However, the workflow automatically clones or updates an unpinned repository and subsequently executes its Python runner without an integrity-verification step; compromise of the upstream repository or its mirror could therefore result in arbitrary code execution with the agent's filesystem and credential access.
Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.
CUE_KEY=$(python3 -c "import json;print(json.load(open('$HOME/.cue/config.json'))['api_key'])" 2>/dev/null || echo "$CUE_API_KEY")
echo "=== 1/4 API Key ===" && [ -n "$CUE_KEY" ] && echo "已配置" || echo "未配置!"
echo "=== 2/4 Cue 服务 ===" && curl -sS --max-time 10 "https://cuecue.cn/api/health" -H "Authorization: Bearer $CUE_KEY"
echo "=== 3/4 搭子 ===" && curl -sS --max-time 10 "https://cuecue.cn/api/playbook" -H "Authorization: Bearer $CUE_KEY" | python3 -c "import sys,json;scenes=json.load(sys.stdin).get('data',{}).get('scenes',[]);buddy=[b for s in scenes for b in s.get('buddies',[]) if b.get('title')=='基金组合穿透与重复配置排雷'];print(f'可用:{len(buddy)}个') if buddy else print('暂不可用')"
echo "=== 4/4 剩余积分 ===" && python3 -c "
import json, urllib.request, os
p = os.path.expanduser('~/.cue/config.json')
The setup explicitly tells users to paste their API key into the conversation instead of configuring it through a secret-input mechanism. This unnecessarily exposes the credential to chat retention, logging, and conversation sharing; the statements that the agent cannot view the key do not prevent that exposure.
This identifies the same credential-handling issue as the SQP-2 finding: users are required to disclose an API key in chat. The exposure is unnecessary for fund analysis and could permit unauthorized API use or consumption of account credits if conversation records are accessed.
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
| 目标格式 | 依赖 | macOS | Ubuntu |
|----------|------|-------|--------|
| Word (.docx) | pandoc | `brew install pandoc` | `sudo apt install pandoc` |
| PDF (.pdf) | pandoc + LaTeX | `brew install --cask basictex` | `sudo apt install texlive-xetex` |
---
Commands invoke sudo or root privileges. Verify this elevated access is necessary and justified.
| 目标格式 | 依赖 | macOS | Ubuntu |
|----------|------|-------|--------|
| Word (.docx) | pandoc | `brew install pandoc` | `sudo apt install pandoc` |
| PDF (.pdf) | pandoc + LaTeX | `brew install --cask basictex` | `sudo apt install texlive-xetex` |
---
The skill mandates referral, purchase, and social-sharing prompts and explicitly prevents other follow-up questions until its promotional workflow is satisfied. Although credit notices are relevant to a paid service, these requirements subordinate the user's analysis task to the provider's marketing objectives; sharing a financial report could also expose sensitive information if the user proceeds without understanding its visibility.
The skill requires the Chinese question and options to be reproduced verbatim and prohibits changing their text. This forces a specific output language without user opt-in; the skill's focus on Chinese funds does not establish a need to force Chinese for account-balance prompts.