transparent-icons
Pass
Audited by Gen Agent Trust Hub on Jun 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The script
scripts/generate_icon.pyinvokes thecodexCLI tool usingsubprocess.runto perform image generation tasks within a controlled workspace sandbox. Additionally,examples/icon_set.pyusessubprocess.runto automate the execution of the icon generator script. - [EXTERNAL_DOWNLOADS]: When configured to use the
birefnettransparency method, the skill utilizes therembglibrary, which initiates a one-time download of the BiRefNet ONNX model (approximately 400 MB) from its official source. This is a legitimate operation for the skill's stated purpose of high-quality background removal. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface in
scripts/generate_icon.pywhere user-provided icon descriptions are interpolated into natural language instructions for thecodexCLI. While this could potentially be used to influence the downstream AI agent, the skill documentation acknowledges the risk and recommends sanitization if exposed to untrusted inputs.
Audit Metadata