Description-Behavior Mismatch
- Category
- MCP Tool Poisoning
- Confidence
- 92% confidence
- Finding
文档转换可以作为生成尽调底稿的辅助步骤,本身并非不当能力。但所提供代码仅包含格式转换,没有展示 Cue 调用、企业信息获取、尽调分析或相关报告内容生成。因此,就该代码块所展示的完整行为而言,其主要用途是格式转换,而不是声明中的企业基本面尽调。未发现额外网络访问或其他敏感操作。
- Content
Security audit
Security checks for vulnerabilities and agentic risk
ClawScan rates this skill suspicious because scanners report unsafe API key handling and workflow instructions that extend beyond its stated due diligence purpose; one local reference was unresolved, so the evidence is incomplete.
Consider the suspicious verdict and the reported API key exposure before installing. Do not paste a live key into chat; use a protected local configuration method and rotate any key already shared. The scan covered the listed component but was marked partial because a local path-like reference could not be resolved, so the results do not establish that all referenced material was reviewed.
文档转换可以作为生成尽调底稿的辅助步骤,本身并非不当能力。但所提供代码仅包含格式转换,没有展示 Cue 调用、企业信息获取、尽调分析或相关报告内容生成。因此,就该代码块所展示的完整行为而言,其主要用途是格式转换,而不是声明中的企业基本面尽调。未发现额外网络访问或其他敏感操作。
The statement that the agent cannot view the key conflicts with the instruction to paste it into the agent conversation and have the agent store it. This misleading assurance can encourage credential disclosure and enable unauthorized API use or credit consumption if the exposed key is obtained by another party; it is the same underlying exposure identified in SQP-2.
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 if s.get('secondary_category')=='信贷尽调' for b in s.get('buddies',[]) if b.get('title')=='企业全景画像'];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 asks the user to paste a live API key into the conversation, unnecessarily exposing it to chat history, model processing, and potentially logging or support systems. This is a genuine credential-handling weakness even though the key belongs to the service legitimately used by the skill; it duplicates the underlying issue in SSD-3.
The skill mandates promotional prompts and gives them priority over other follow-up questions, extending the workflow beyond the user's research request. This is an output and workflow-integrity issue, although the balance condition and explicit user choice limit the risk: the instructions do not automatically purchase credits or publish a report.
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 requires the low-balance question and options to be reproduced verbatim, explicitly prohibiting changes. This forces Chinese user-facing output without offering a language choice or allowing adaptation to the user's language.