docx-reader
Pass
Audited by Gen Agent Trust Hub on Sep 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill operates by instructing the agent to execute shell commands in a WSL environment (e.g.,
wsl python3 scripts/read_docx.py "/mnt/c/path/to/file.docx"). This pattern relies on the agent's ability to construct and run command-line tools to process local files. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external Microsoft Word (.docx) files, which serves as a vector for indirect prompt injection.
- Ingestion points: Text and table data are extracted from user-provided files in
scripts/read_docx.pyusing thepython-docxlibrary. - Boundary markers: The skill body in
SKILL.mdlacks clear delimiters or instructions to the agent to treat the extracted document content as untrusted data or to ignore any commands found within it. - Capability inventory: The skill environment allows for shell command execution (
wsl) and file system operations (writing Markdown files), creating a significant impact if an injection attack succeeds. - Sanitization: There is no evidence of text sanitization, instruction filtering, or escaping in the Python script before the content is returned to the agent's context.
Audit Metadata