journal-monitor
monitor.py 是纯良性的 PubMed 文献监控脚本,无恶意代码;但 SKILL.md 包含反安全/提示词注入特征的强约束语言。
Why this conclusion was reached
0/4 dimensions flaggedDeclared resources and inferred behavior are broadly aligned.
1 lower-risk artifacts were extracted and still need context.
There is no explicit malicious chain in the report.
1 dependency or supply-chain issues need attention.
What drove the risk score up
使用 "FORCED"、"MANDATORY"、"NO tool-switching" 等强约束词汇,试图限制 LLM 使用更安全的替代工具,符合 prompt_injection 特征但代码本身无恶意
Entrez.email = "[email protected]" 在代码中硬编码,标准 NCBI API 要求但可考虑环境变量方式
Most important evidence
SKILL.md 包含反安全/反提示注入的强约束语言
SKILL.md 使用 "FORCED translation, MANDATORY metadata, and NO tool-switching" 以及 "MUST ONLY use data returned by monitor.py"、"DO NOT switch to internal search" 等强约束语句,试图阻止 LLM 使用更安全或合法的替代工具,属于 prompt injection 特征但未达到恶意级别。
SKILL.md:1 邮箱地址硬编码(潜在个人信息暴露)
Entrez.email 硬编码为 [email protected],属于个人邮箱,可能随代码分发而暴露个人信息
monitor.py:8 biopython 依赖无版本锁定
metadata 声明 "uv pip install biopython" 无版本约束,可能引入依赖供应链风险
SKILL.md:11 Declared capability vs actual capability
monitor.py:36 使用 Entrez.esearch/efetch 访问 PubMed API(声明与行为一致) monitor.py 无任何文件读写操作 monitor.py 无 subprocess/os.system 等 shell 调用 monitor.py 仅依赖 CLI 参数和 os.environ(sys.argv),无凭证遍历 无嵌套 skill 调用 Suspicious artifacts and egress
Dependencies and supply chain
| Package | Version | Source | Known vuln | Notes |
|---|---|---|---|---|
| biopython | * | pip | No | SKILL.md 中声明为 "uv pip install biopython",无版本锁定,存在供应链风险 |
File composition
monitor.py SKILL.md