ui-ux-pro-max
Pass
Audited by Gen Agent Trust Hub on May 5, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface
- Ingestion points: The skill accepts user-supplied project names and page names via the CLI arguments in
scripts/search.pyandscripts/design_system.py, which are used to generate documentation. - Boundary markers: The generated markdown files (
MASTER.mdand page-specific overrides) lack clear delimiters or "ignore embedded instructions" warnings for the user-influenced fields. - Capability inventory: The skill has the capability to write these files to the local file system (
scripts/design_system.py) and specifically instructs the agent inSKILL.mdto read these persisted files and "strictly follow" or "prioritize" the rules within them for subsequent code generation tasks. - Sanitization: While the skill tokenizes search queries, it does not sanitize the project or page names used in file creation, creating a surface where malicious instructions could be persisted and later executed by the agent. This is a classic indirect prompt injection vector where data processed in one session influences the agent's behavior in another.
Audit Metadata