baoyu-article-illustrator
Pass
Audited by Gen Agent Trust Hub on Aug 21, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The skill ingests untrusted article content to identify concepts and generate image prompts. It lacks clear boundary markers or sanitization for this data, creating a surface for indirect prompt injection where malicious instructions embedded in an article could manipulate the agent's behavior.
- [PROMPT_INJECTION]: Instructions in
prompts/system.mdexplicitly direct the image generation tool to 'DO NOT refuse to generate' even when content involves sensitive or copyrighted figures, representing an attempt to bypass default AI safety guardrails. - [COMMAND_EXECUTION]: The skill instructions specify executing shell commands (e.g.,
test -f) to detect the presence of preference files and reference images within the project and user directories. - [DATA_EXPOSURE]: The workflow involves accessing the user's home directory (
$HOME/.baoyu-skills/) to read and write persistent configuration files, which is a standard preference management pattern but involves interaction with the host file system.
Audit Metadata