Back to skill

Security audit

B站视频全解析

Security checks for vulnerabilities and agentic risk

Overview

This skill needs review because it repeatedly asks for your API key in chat and requires promotional follow-ups.

Configure the API key locally instead of sending it in chat, and rotate any key already shared. Expect paid processing, locally saved transcripts or media, and referral/top-up prompts that you may decline. Remote retention guarantees and the optional Bridge remain unverified. The supplied static scan reported no findings, but the credential-handling instructions warrant review.

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
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (9)

Onboarding instructions require users to disclose their complete API key in chat under the pretext of convenient local configuration.

High
Category
Data Exfiltration
Confidence
99% confidence
Finding

Repeated onboarding instructions encourage disclosure of the entire service credential through chat, even though local configuration alternatives already exist. Anyone obtaining a retained transcript could potentially use the key for unauthorized service access or credit consumption; this is the same underlying credential-exposure issue as SDI-4, not an additional independent compromise.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

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.

Content

Scanner excerpt · SKILL.md (reported line 64)May include surrounding context.

方式一:环境变量(写入 shell 启动文件后重开终端/客户端生效)

export CUE_API_KEY="sk你的key"

方式二:本地配置文件

mkdir -p ~/.cue && chmod 700 ~/.cue && echo '{"api_key": "sk你的key"}' > ~/.cue/config.json && chmod 600 ~/.cue/config.json

text

- **远程端点直连(无需本地 Bridge、无需配置 MCP)**:B 站 URL 解析只走远程端点 `https://mcp.cuecue.cn/api/omni-reader/mcp/`(streamable-http),鉴权头 `Authorization: Bearer <Cue Key>`。不依赖 `~/.mcp.json` 注册 MCP,也不装本地 Bridge;本地 Bridge 仅在降级方案中使用。

The setup workflow solicits API keys in conversation despite the documented local-only handling assurance.

Medium
Category
Excessive Agency
Confidence
99% confidence
Finding

The setup explicitly asks users to paste a complete API key into the conversation, exposing it to the chat-processing infrastructure and potentially retained transcripts before it is written locally. Instructions not to echo the key cannot prevent this initial disclosure, and the local-only assurance may give users a misleading impression of confidentiality.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 91)May include surrounding context.

md
CUE_KEY=$(python3 -c "import json,os;p=os.path.expanduser('~/.cue/config.json');print(json.load(open(p)).get('api_key','') if os.path.exists(p) else '')" 2>/dev/null || true)
CUE_KEY=${CUE_KEY:-$CUE_API_KEY}
echo "=== 1/4 API Key ===" && { [ -n "$CUE_KEY" ] && echo "已配置" || echo "未配置!"; }
echo "=== 2/4 远程端点 ===" && curl -sS --max-time 10 -X POST "https://mcp.cuecue.cn/api/omni-reader/mcp/" \
  -H "Authorization: Bearer $CUE_KEY" \
  -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}' \

The skill mandates referral and purchase promotion after successful tasks, beyond its video-processing purpose.

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

The skill mandates referral and purchase messaging and explicitly displaces other follow-up questions, allowing provider marketing requirements to override the user's task-oriented interaction. This is a limited workflow-integrity issue rather than evidence of payment theft: the supplied instructions do not authorize automatic purchases or establish that the payment links are fraudulent.

Content

No source excerpt is available for this finding.

The mandatory low-balance prompt fixes user-facing output in Chinese without offering a language choice.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

L203 requires the prompt title and options to be used verbatim without modification, and L205–L209 supply exclusively Chinese text. This mandates Chinese output when the balance condition is met, without user language opt-in or a documented justification for that language restriction.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
60% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 421)May include surrounding context.

解析调用手册

curl 三步

bash
# ① 提交解析(记下返回的 operation_id)

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 579)May include surrounding context.

bash
# ① 取 cid
curl -sS --compressed "https://api.bilibili.com/x/web-interface/view?bvid=BVxxxxxxxx" \
  -H "User-Agent: Mozilla/5.0" -H "Referer: https://www.bilibili.com/"
# ② 取免登录 360p MP4 直链(durl[0].url,带 deadline 时效参数,需尽快提交)
curl -sS --compressed "https://api.bilibili.com/x/player/playurl?bvid=BVxxxxxxxx&cid=<cid>&qn=64&platform=html5&high_quality=1" \

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 582)May include surrounding context.

curl -sS --compressed "https://api.bilibili.com/x/web-interface/view?bvid=BVxxxxxxxx"
-H "User-Agent: Mozilla/5.0" -H "Referer: https://www.bilibili.com/"

② 取免登录 360p MP4 直链(durl[0].url,带 deadline 时效参数,需尽快提交)

curl -sS --compressed "https://api.bilibili.com/x/player/playurl?bvid=BVxxxxxxxx&cid=&qn=64&platform=html5&high_quality=1"
-H "User-Agent: Mozilla/5.0" -H "Referer: https://www.bilibili.com/"

③ 把 MP4 直链作为 source 提交远程端点 parse(detail:"text"),流程同标准工作流

text