swiftui-pro
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEPROMPT_INJECTIONMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The instructions include a 'hallucinated context' directive stating that 'iOS 26 exists' and is the default target. This technique is used to override the model's internal knowledge of current software versions to force specific behavior, a pattern often seen in safety bypass attempts.- [METADATA_POISONING]: The metadata identifies the author as a well-known third-party expert rather than the actual skill provider. This could lead users to misplace trust in the skill's recommendations based on a perceived authoritative source.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted Swift code provided by users, which is a potential vector for indirect injection.
- Ingestion points: The skill reads and reviews Swift source code files provided by the user (SKILL.md).
- Boundary markers: There are no instructions to use delimiters or ignore instructions that might be embedded in code comments within the files being reviewed.
- Capability inventory: The skill is limited to code review and reporting; no dangerous file-writing, network operations, or shell commands are invoked in the instructions.
- Sanitization: The skill lacks any explicit instructions to sanitize or escape the content of the code files before analysis.
Audit Metadata