design-protocol
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes project-specific design protocol files (
design/protocol/*.json) to guide the agent's UI generation decisions. While these files are validated structurally, the descriptive content within them enters the agent's context. - Ingestion points: JSON files defining design layers (L1–L7) located in the project's protocol directory.
- Boundary markers: The skill instructions mandate explicit citations (e.g.,
protocol:L2/button) for every design decision, which creates a logical trace for generated content. - Capability inventory: The skill's scope is limited to UI code generation; it does not request network access or sensitive file write capabilities.
- Sanitization: The skill uses a local script to validate the structure of all protocol files against predefined JSON schemas.
- [COMMAND_EXECUTION]: The skill instructions direct the agent to run a local Python utility (
scripts/validate.py) to validate design protocol files. This script is part of the skill package and operates on local file paths.
Audit Metadata