image-ocr
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes external binaries and subprocesses for OCR tasks. It uses the tesseract CLI and can launch a separate Python process for the MLX backend using the --mlx-worker-python argument. This is standard behavior for interfacing with specialized OCR engines.
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted image data and outputs the extracted text into a Markdown file. If an image contains malicious instructions, they could be reflected in the final document and potentially influence future agent actions.
- Ingestion points: Images located in the user-provided directory specified via the --input argument in SKILL.md.
- Boundary markers: None explicitly mentioned in the documentation.
- Capability inventory: Reads image files, writes to a Markdown file, and executes subprocesses for Tesseract and MLX backends.
- Sanitization: No explicit sanitization or filtering of the extracted OCR text is described.
Audit Metadata