ui-ux-pro-max
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPRIVILEGE_ESCALATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user-supplied queries through a CLI search tool to retrieve UI guidelines, creating a natural attack surface for indirect prompt injection.
- Ingestion points: The
queryargument passed toscripts/search.py. - Boundary markers: Output is structured using ASCII boxes (default) or Markdown blocks, which helps delimit data from instructions.
- Capability inventory: The Python scripts (
core.py,design_system.py,search.py) perform local CSV data retrieval and console printing. They do not possess capabilities for network operations, file writing, or arbitrary shell execution. - Sanitization: The
BM25search engine incore.pytokenizes input by stripping non-alphanumeric characters and converting text to lowercase before processing, providing a baseline layer of input sanitization. - [COMMAND_EXECUTION]: The skill is designed to be used by executing local Python scripts to query the design database.
- Evidence: Instructions in
SKILL.mdguide the agent to runpython3 .claude/skills/ui-ux-pro-max/scripts/search.pywith user-derived keywords. - Context: This behavior is the primary intended function of the skill and the scripts are limited to data retrieval from the provided CSV files.
- [EXTERNAL_DOWNLOADS]: The skill references external assets for typography and design implementation.
- Evidence:
data/typography.csvcontains numerous links to the Google Fonts service (fonts.googleapis.comandfonts.google.com). - Context: These references target a well-known service for legitimate design purposes and do not involve executable code downloads.
- [PRIVILEGE_ESCALATION]: The documentation provides standard environment setup instructions that involve elevated permissions.
- Evidence:
SKILL.mdincludes a prerequisite section for installing Python usingsudo apt install python3on Debian-based systems. - Context: These are standard, transparent instructions for a user to configure the required runtime environment and are not hidden or executed silently by the skill itself.
Audit Metadata