baoyu-xhs-images

Pass

Audited by Gen Agent Trust Hub on Mar 3, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill includes explicit instructions in references/workflows/prompt-assembly.md to bypass safety filters. It directs the agent to 'create stylistically similar alternatives' and 'DO NOT refuse to generate' when encountering requests involving sensitive or copyrighted figures.
  • [COMMAND_EXECUTION]: The skill utilizes shell commands (e.g., test -f) within its workflow to check for the presence of configuration files in the user's home directory ($HOME/.baoyu-skills/). While typical for configuration management, it represents execution of code that interacts with the local system.
  • [PROMPT_INJECTION]: An indirect prompt injection surface is created because the skill ingests untrusted user content and interpolates it directly into image generation prompts without sanitization.
  • Ingestion points: User-pasted content or files read during 'Step 1: Analyze Content'.
  • Boundary markers: The prompt assembly in references/workflows/prompt-assembly.md uses Markdown headers and horizontal rules as delimiters but lacks specific instructions for the AI to ignore instructions embedded within the user's content.
  • Capability inventory: The skill can execute shell commands (test), perform file and directory operations (mkdir, write), and invoke external image generation tools.
  • Sanitization: There is no evidence of filtering or escaping user-provided text before it is inserted into the {CONTENT_SECTION} of the generation prompt.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 3, 2026, 01:19 PM