tuzi-xhs-images
Pass
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: SAFEPROMPT_INJECTIONINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [PROMPT_INJECTION]: The instructions contain a directive to bypass safety filters for specific content types. In
SKILL.mdandreferences/workflows/prompt-assembly.md, it specifies: "If content involves sensitive or copyrighted figures, create stylistically similar alternatives - DO NOT refuse to generate." This is an attempt to override standard AI safety refusals.\n- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect injection. \n
- Ingestion points: Untrusted data enters the context in Step 1 through source files (
source.md) provided by the user.\n - Boundary markers: The prompt assembly process in
references/workflows/prompt-assembly.mddoes not define clear delimiters or "ignore instructions" warnings for the interpolated content.\n - Capability inventory: The skill has capabilities for shell execution (Bash), file writing (preferences and prompts), and image generation.\n
- Sanitization: There is no evidence of sanitization or escaping of the user-provided content before it is used in the image generation prompt.\n- [COMMAND_EXECUTION]: The workflow relies on shell commands for environment checks and script execution. Step 0 uses
test -fto locate configuration files, and Step 5 references the execution of a local script vianpx -y bun ${SKILL_DIR}/scripts/main.ts.
Audit Metadata