web-interface-architect
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill consists of instructional text for the agent to act as a frontend architect. It does not provide any scripts, executable code, or tool definitions, and performs no network or file system operations.
- [METADATA_POISONING]: The skill's frontmatter incorrectly identifies the license as 'Proprietary', while the accompanying LICENSE.txt file specifies the Apache License 2.0. This is a metadata discrepancy with no impact on security.
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user requirements for UI generation. 1. Ingestion points: User requirements provided as input (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: None; the skill defines no tools or subprocess capabilities. 4. Sanitization: Absent. Given the lack of capabilities, this does not pose a significant risk.
Audit Metadata