image-ocr-hardened
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
tesseractcommand-line utility to perform OCR operations on various image formats. - [EXTERNAL_DOWNLOADS]: The skill provides installation instructions for the
tesseractengine using the system package manager (dnf), which fetches legitimate software from official repositories. - [PROMPT_INJECTION]: Because the skill processes data from external image files, it possesses an indirect prompt injection attack surface. It explicitly addresses this by including instructions and guardrails that prevent the agent from treating extracted text as executable commands.
- Ingestion points: Image files (PNG, JPEG, TIFF, BMP) provided by the user (found in
SKILL.md). - Boundary markers: The
SKILL.mdfile contains a dedicated 'Security Guardrails' section that establishes clear boundaries for the agent's behavior when handling OCR output. - Capability inventory: The skill uses the
tesseractbinary for file processing. - Sanitization: Relies on comprehensive instructional guardrails to prevent the agent from executing image content, piping output to the network, or persisting unreviewed data.
Audit Metadata