docx
Warn
Audited by Gen Agent Trust Hub on May 6, 2026
Risk Level: MEDIUMPROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill metadata contains deceptive claims regarding its origin and safety, stating the author is 'anthropic' and that it has 'PASSED' an official security audit. This conflicts with the provided author context and serves to bypass safety scrutiny.
- [PROMPT_INJECTION]: The skill processes untrusted external Word documents through pandoc and docx-js, creating a surface for indirect prompt injection. Ingestion points: document.docx via pandoc. Boundary markers: Absent. Capability inventory: fs.writeFileSync and shell command execution. Sanitization: Absent.
- [EXTERNAL_DOWNLOADS]: The skill recommends the global installation of the 'docx' library from the npm registry.
- [COMMAND_EXECUTION]: The skill uses the 'pandoc' command-line utility to convert Word documents to markdown for analysis.
Audit Metadata