stitch-ui-prompt-architect
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from project files and user input to construct prompts for UI generation, creating an attack surface where malicious data could influence agent behavior.
- Ingestion points: The skill reads
DESIGN.md(specifically Section 6) from the project environment and ingests rawuserRequeststrings and Design Spec JSON objects. - Boundary markers: There are no explicit delimiters or instructions to treat the ingested data as untrusted content or to disregard embedded commands within the
DESIGN.mdcontent. - Capability inventory: The skill uses the
Readtool to access filesystem content. The resulting prompts are intended to be passed to a screen generation tool, which constitutes a multi-step capability chain. - Sanitization: The skill performs no validation, escaping, or sanitization on the data extracted from
DESIGN.mdor the user request before interpolating it into the final prompt structure.
Audit Metadata