tesseract

Pass

Audited by Gen Agent Trust Hub on Sep 21, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill demonstrates processing external image and PDF files for text extraction.
  • Ingestion points: Data enters the context via standard library calls like Image.open() and convert_from_path() in the Python snippets within SKILL.md.
  • Boundary markers: The provided examples are educational and do not include instructions for the agent to ignore content within the images, which is typical for code snippets.
  • Capability inventory: The skill only demonstrates printing the extracted text to the console. It lacks any dangerous capabilities such as executing the extracted text or sending it over the network.
  • Sanitization: Standard for example code, no extra sanitization is shown, but the lack of follow-up capabilities makes this surface low risk.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 21, 2026, 01:04 PM
Security Audit — agent-trust-hub — tesseract