Security audit
Security checks for vulnerabilities and agentic risk
This CSV skill is mostly a local data-processing guide, but it asks for broad read/write/exec authority and includes unclear non-CSV, callback, and API-key language that users should review before installing.
Install only if you intend to use it for CSV files, and review each operation before allowing writes or command execution. Avoid providing secrets or callback URLs unless the publisher clarifies why they are needed, and direct exports or split outputs to a known folder to prevent unintended copies of sensitive business or financial data.
The description says the skill should be used for broad file-processing, document-conversion, and content-extraction tasks, far beyond CSV handling. Overly broad invocation guidance can cause agents to route unrelated files or sensitive data into a skill with read, write, and exec capabilities, increasing the chance of unintended file access or destructive transformations.
The description states '支持中文交互' in a way that indicates a fixed language expectation, and the document does not offer a language/locale choice or explain a justified region-specific limitation. This can violate language/locale policy requirements when the skill is presented for broad multi-agent use.
The input schema declares a callback_url for asynchronous completion notifications, implying the skill can make outbound callbacks when processing finishes. Elsewhere, the document only describes local CSV parsing, cleaning, merging, splitting, and exporting operations; no code-facing behavior or examples implement notification delivery, so the documentation overstates behavior.
The skill documents split operations that create multiple files but does not prominently warn users that running the operation writes many outputs to disk. In an agent context, silent file creation can surprise users, overwrite storage workflows, expose sensitive subsets into new locations, or amplify data sprawl when processing confidential CSVs.
The export feature rewrites data to disk with potentially different encoding, delimiter, cleaned column names, and transformed types, but the description does not clearly warn about these persistent changes. This can lead to integrity issues, accidental overwrites, or unintended disclosure if users assume the operation is only in-memory analysis.
This line makes a concrete product claim that the skill has built-in security checks. However, the surrounding documentation describes CSV parsing, cleaning, type conversion, and export, while the security section provides recommendations rather than implemented checks, creating an intent-versus-behavior contradiction.
No suspicious patterns detected.
