adversarial-swift-ui
Pass
Audited by Gen Agent Trust Hub on Jul 20, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data in the form of SwiftUI source code or diffs provided in the
{{TARGET_CONTENT_OR_PATHS}}placeholder withinreferences/reviewer-prompt.md. This data is directly interpolated into the prompt for a subagent. - Ingestion points: The
reviewer-prompt.mdtemplate accepts external code content via{{TARGET_CONTENT_OR_PATHS}}and a description via{{TARGET_DESCRIPTION}}. - Boundary markers: While the prompt template structures the input as a 'Review Target', it lacks explicit boundary markers or instructions to the subagent to ignore any instructions found within the code being analyzed.
- Capability inventory: The skill dispatches a 'Task subagent' (via
SKILL.mdStep 4) which has the capability to render a security or architectural verdict. - Sanitization: There is no evidence of escaping or sanitizing the input code before interpolation into the subagent prompt.
Audit Metadata