docx-to-knowledge
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The Python scripts (
extract_final.py,gen_html_v4.py,refine_captions.py) utilize hardcoded absolute file paths belonging to the author's local machine (e.g.,/Users/fred/Documents/GitHub/cycleuser/...). These scripts perform file system operations such as reading, writing, and directory creation at these specific paths. Additionally, the skill documentation mentions running automated quality checks using tools like Playwright and static analysis vianode --check, which involves executing system commands. - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external
.docxfiles, creating a surface for indirect prompt injection. Maliciously crafted content within these documents could potentially influence the agent during the extraction and cleaning process. - Ingestion points: External
.docxfiles processed byscripts/extract_final.py. - Boundary markers: None identified in the processing logic to distinguish between data and instructions.
- Capability inventory: File system read/write access, directory creation, and the ability to execute shell commands for verification.
- Sanitization: Basic text normalization is implemented, but there is no specific sanitization for preventing prompt injection from within the ingested document content.
Audit Metadata