draft-ux
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository, such as
requirements.mdandui-brief.md, to generate code and documentation. This creates an attack surface where malicious instructions in these files could potentially influence the agent's behavior. - Ingestion points: The agent reads
requirements.mdandui-brief.mdin Steps 1, 2, and 5 to identify UX moments and constraints. - Boundary markers: There are no explicit instructions for the agent to treat input from these files as untrusted or to ignore embedded instructions.
- Capability inventory: The skill allows the agent to modify code files (
app.js,index.html) and delete files in the repository during the prototype cleanup stage (Step 6). - Sanitization: The skill lacks steps for validating or sanitizing the content extracted from project files before it is used in code generation or documentation updates.
Audit Metadata