Security audit
Security checks for vulnerabilities and agentic risk
The skill is a CSV helper, but it asks for command and write capabilities and describes API/key and command-execution behavior that is broader than its CSV purpose.
Install only if you are comfortable with a CSV skill that may be invoked broadly and that advertises command execution, file writing, and API/key handling beyond simple local CSV processing. Keep it limited to non-sensitive CSVs unless you can review and constrain the agent's tool access, output paths, overwrite behavior, and any external API use.
Documenting system command execution for a CSV-processing skill is dangerous because command execution is a high-risk primitive not justified by the stated business purpose. In an agent setting, this can normalize or authorize running shell commands on attacker-controlled filenames, paths, or derived content, enabling arbitrary code execution or sensitive file access.
The manifest description states the skill '支持基础合并与导出' ('supports basic merging and export'), implying merge support is available. Later sections repeatedly state that multi-file merge (merge_csvs) is a paid-only feature and that the free version does not support multi-file merging, creating an active contradiction about intended behavior.
The trigger description is overly broad, covering generic data analysis, reporting, statistics, and visualization scenarios rather than narrowly scoping to safe CSV handling. In an agent ecosystem, broad activation criteria can cause the skill to be selected for unrelated tasks, increasing exposure of dangerous capabilities like write access or command execution in contexts where users did not intend them.
L003 明确写有“支持中文交互，无需复杂配置即开即用”，整体文档也仅提供中文使用说明，但没有说明这是可选语言还是区域限定要求。根据语言/地区策略，若技能默认或强制单一语言而未提供用户选择或正当限定，属于自然语言策略风险。
The skill documents export and file-writing behavior without clearly warning users about filesystem side effects such as overwriting outputs, creating new files, or storing cleaned data on disk. In agent workflows, insufficient disclosure around writes can lead to unintended data modification, accidental disclosure into shared directories, or destructive overwrites.
A CSV handler has no clear need to require API key setup or imply external API usage, yet the documentation introduces both. This expands the trust boundary unnecessarily and may encourage users or agents to expose secrets to a skill that should be purely local, increasing the risk of credential leakage or unjustified outbound data flows.
The skill is presented as a CSV-processing utility, but later advertises generic API integration and system command execution capabilities unrelated to that narrow purpose. This capability mismatch increases the chance that the skill is invoked in broader contexts and then granted dangerous execution paths that could access local files, run commands, or exfiltrate data under the guise of CSV handling.
