Description-Behavior Mismatch
- Category
- MCP Tool Poisoning
- Confidence
- 99% confidence
- Finding
文档转换可以作为报告流程的辅助实现,因此不单独将 Word/PDF 导出认定为违规能力。但就所提供的代码而言,其全部行为都是 Markdown 文档转换,没有实现声明中的核心扫描、分析及交付功能,故判定描述与该代码块的实际行为不匹配。此结论仅针对所提供代码,不推断未展示的实现。
- Content
Security audit
Security checks for vulnerabilities and agentic risk
ClawScan marks this skill suspicious with medium confidence because scanners report a mismatch between its market-research description and the provided document-conversion code, along with remote fetching and instructions for scheduled runs.
Consider the suspicious verdict before installing. Review the mismatch between the promised research features and the inspected code, avoid pasting an API key into chat, and consider whether you want the skill to create recurring scheduled runs or promote purchases and referrals. Scanning was partial because a referenced local path was missing, so the evidence is incomplete.
文档转换可以作为报告流程的辅助实现,因此不单独将 Word/PDF 导出认定为违规能力。但就所提供的代码而言,其全部行为都是 Markdown 文档转换,没有实现声明中的核心扫描、分析及交付功能,故判定描述与该代码块的实际行为不匹配。此结论仅针对所提供代码,不推断未展示的实现。
YARA rule matched a known malware signature (reverse shell, backdoor, ransomware, C2 framework, or info stealer).
*触发时间、输出目录、如何关掉**(Claude Code 删任务 / `launchctl unload` / `crontab -e` 删除对应行)。
### 必须提前告知的三个坑
- **每次运行都会消耗 Cue 积分**;设多个时点按次数叠加,余额不足时当天会失败,需按「积分不足专项提示」补充积分,不会自动跳过。
- **无界面跑批出不全两张图**:launchd / crontab 只能自动出报告 md + 报告长图;看板图需要 agent 先重排 SOP JSON,须下次交互时基于已落盘的 `.md` 补生成。Claude Code 的 `CronCreate` 触发时会叫醒 agent,可完整出两张图。
- **法定节假日仍会触发**:cron 只认周一到周五;节假日可能返回空结果或沿用前一交易日数据,属正常现象。报告与图片只落盘 `~/cue-reports/`,不会推送,需要自己打开看。
---
## 环境要求
Runner 来源:[GitHub - sensedeal/cue-skills](https://github.com/sensedeal/cue-skills)([Gitee 镜像](https://gite
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')=='短线盘面' for b in s.get('buddies',[]) if b.get('title')=='24h热点与催化剂追踪'];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')
Without declared permissions the skill's intent is opaque and cannot be validated.
The setup explicitly instructs users to paste an API credential into chat, exposing it to conversation storage, logging, and any parties with transcript access. The legitimate need to authenticate Cue does not require sending the secret through the assistant's conversational context, and the claimed prohibition on viewing keys does not prevent this exposure.
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
| 环境 | 配置方式 |
|------|---------|
| Claude Code | 用 `CronCreate`,cron 表达式 `0 10 * * 1-5`(自定义时间按用户所选取 分 时 日 月 周);触发时由 agent 完整执行「跑研究 → 出两张图 → 交付」,与手动跑一致 |
| macOS | 建 `~/Library/LaunchAgents/com.cue.hotspot-tracker.plist`,`StartCalendarInterval` 设 `Weekday` 1–5、`Hour` 10、`Minute` 0,再 `launchctl load`;只跑上面这条 shell 链路(报告 + 长图) |
| Linux | `crontab -e` 追加 `0 10 * * 1-5 <上面的脚本>` |
配完必须回报三件事:**触发时间、输出目录、如何关掉**(Claude Code 删任务 / `launchctl unload` / `crontab -e` 删除对应行)。
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
| 环境 | 配置方式 |
|------|---------|
| Claude Code | 用 `CronCreate`,cron 表达式 `0 10 * * 1-5`(自定义时间按用户所选取 分 时 日 月 周);触发时由 agent 完整执行「跑研究 → 出两张图 → 交付」,与手动跑一致 |
| macOS | 建 `~/Library/LaunchAgents/com.cue.hotspot-tracker.plist`,`StartCalendarInterval` 设 `Weekday` 1–5、`Hour` 10、`Minute` 0,再 `launchctl load`;只跑上面这条 shell 链路(报告 + 长图) |
| Linux | `crontab -e` 追加 `0 10 * * 1-5 <上面的脚本>` |
配完必须回报三件事:**触发时间、输出目录、如何关掉**(Claude Code 删任务 / `launchctl unload` / `crontab -e` 删除对应行)。
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
|------|---------|
| Claude Code | 用 `CronCreate`,cron 表达式 `0 10 * * 1-5`(自定义时间按用户所选取 分 时 日 月 周);触发时由 agent 完整执行「跑研究 → 出两张图 → 交付」,与手动跑一致 |
| macOS | 建 `~/Library/LaunchAgents/com.cue.hotspot-tracker.plist`,`StartCalendarInterval` 设 `Weekday` 1–5、`Hour` 10、`Minute` 0,再 `launchctl load`;只跑上面这条 shell 链路(报告 + 长图) |
| Linux | `crontab -e` 追加 `0 10 * * 1-5 <上面的脚本>` |
配完必须回报三件事:**触发时间、输出目录、如何关掉**(Claude Code 删任务 / `launchctl unload` / `crontab -e` 删除对应行)。
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
| macOS | 建 `~/Library/LaunchAgents/com.cue.hotspot-tracker.plist`,`StartCalendarInterval` 设 `Weekday` 1–5、`Hour` 10、`Minute` 0,再 `launchctl load`;只跑上面这条 shell 链路(报告 + 长图) |
| Linux | `crontab -e` 追加 `0 10 * * 1-5 <上面的脚本>` |
配完必须回报三件事:**触发时间、输出目录、如何关掉**(Claude Code 删任务 / `launchctl unload` / `crontab -e` 删除对应行)。
### 必须提前告知的三个坑
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
### 必须提前告知的三个坑
- **每次运行都会消耗 Cue 积分**;设多个时点按次数叠加,余额不足时当天会失败,需按「积分不足专项提示」补充积分,不会自动跳过。
- **无界面跑批出不全两张图**:launchd / crontab 只能自动出报告 md + 报告长图;看板图需要 agent 先重排 SOP JSON,须下次交互时基于已落盘的 `.md` 补生成。Claude Code 的 `CronCreate` 触发时会叫醒 agent,可完整出两张图。
- **法定节假日仍会触发**:cron 只认周一到周五;节假日可能返回空结果或沿用前一交易日数据,属正常现象。报告与图片只落盘 `~/cue-reports/`,不会推送,需要自己打开看。
---
Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.
### 必须提前告知的三个坑
- **每次运行都会消耗 Cue 积分**;设多个时点按次数叠加,余额不足时当天会失败,需按「积分不足专项提示」补充积分,不会自动跳过。
- **无界面跑批出不全两张图**:launchd / crontab 只能自动出报告 md + 报告长图;看板图需要 agent 先重排 SOP JSON,须下次交互时基于已落盘的 `.md` 补生成。Claude Code 的 `CronCreate` 触发时会叫醒 agent,可完整出两张图。
- **法定节假日仍会触发**:cron 只认周一到周五;节假日可能返回空结果或沿用前一交易日数据,属正常现象。报告与图片只落盘 `~/cue-reports/`,不会推送,需要自己打开看。
---
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` |
---
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
def _chrome(exe, args):
return subprocess.run([exe, "--headless=new", "--disable-gpu", *args],
check=True, capture_output=True, text=True)
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
_chrome(chrome, ["--hide-scrollbars", "--force-device-scale-factor=2",
f"--window-size={PAGE_WIDTH},{h + 4}",
f"--screenshot={raw}", html_path.as_uri()])
subprocess.run(["sips", "--resampleWidth", str(width), str(raw),
"--out", str(png_path)], check=True, capture_output=True)
raw.unlink(missing_ok=True)
else:
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
_chrome(chrome, ["--hide-scrollbars", "--force-device-scale-factor=2",
f"--window-size={PAGE_WIDTH},{height + 4}",
f"--screenshot={raw}", html_path.as_uri()])
subprocess.run(["sips", "--resampleWidth", str(width), str(raw),
"--out", str(png_path)], check=True, capture_output=True)
raw.unlink(missing_ok=True)
else:
subprocess module calls execute external commands. Without careful input validation, this enables command injection.
def _chrome(exe, args):
return subprocess.run([exe, "--headless=new", "--disable-gpu", *args],
check=True, capture_output=True, text=True)
The skill makes referral, purchase, and social-sharing promotion a mandatory continuation of a market-research task when credits are low. Although it does not automatically purchase or publish anything, it redirects the assistant toward undisclosed commercial objectives and encourages sharing a report identifier outside the requested research workflow.