byted-mediakit-image

Pass

Audited by Gen Agent Trust Hub on Nov 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses the 'shell' permission to execute the 'mediakit-cli' tool. It passes user-provided URLs or local file paths as arguments to this command-line interface.
  • [DATA_EXPOSURE]: Multiple reference files (e.g., 'image-ocr.md', 'add-image-watermark.md', 'probe-image-metadata.md') explicitly state that the '--image-url' parameter supports local file paths. This grants the skill the capability to read images directly from the agent's host filesystem.
  • [INDIRECT_PROMPT_INJECTION]: The 'image-ocr' tool (detailed in 'reference/image-ocr.md') extracts text from images and returns the content, coordinates, and confidence to the agent. This content is then processed by the LLM, creating a potential surface for indirect prompt injection if an image contains malicious instructions.
  • Ingestion points: The '--image-url' parameter in 'reference/image-ocr.md' and other tools that process image content.
  • Boundary markers: No specific delimiters or instructions to ignore embedded text are provided in the skill guidelines.
  • Capability inventory: The skill can execute shell commands via 'mediakit-cli' to process media.
  • Sanitization: There is no evidence of filtering or sanitizing the extracted OCR text before it is returned to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 5, 2026, 09:02 AM