ck:frontend-design
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documentation includes numerous examples of shell commands used to execute local Python scripts (
scripts/gemini_batch_process.pyandscripts/media_optimizer.py) for processing visual assets and interacting with AI APIs. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external media, such as screenshots, videos, and competitor website references, through multimodal AI models, which creates a potential surface for adversarial instructions embedded in visual data. • Ingestion points: Multimodal analysis of user-provided images, videos, and captured web content. • Boundary markers: The instructions do not specify explicit delimiters or isolation protocols for processing untrusted media. • Capability inventory: The skill has capabilities for file system writing, frontend code generation (React, Vue, Three.js), and sub-agent orchestration. • Sanitization: The workflow relies on the safety guardrails of the underlying AI models (Gemini/Imagen).
- [REMOTE_CODE_EXECUTION]: The skill references and executes scripts within a 'scripts/' directory that were not included in the provided file set, meaning their logic and safety cannot be directly verified through this analysis.
Audit Metadata