do-review-ios
Warn
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: MEDIUMMETADATA_POISONINGDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: Documentation in files such as references/ios-animation-implementation/SKILL.md and references/ios-animation-design/SKILL.md contains demonstrably false and misleading information, claiming to support non-existent technologies like iOS 26 and Liquid Glass. This deceptive content can cause the agent to provide incorrect or hallucinated advice.
- [DYNAMIC_EXECUTION]: The skill instructs the agent to execute swift build and swift test on the repository being reviewed. This represents a significant execution surface, as a malicious repository could contain code that executes during the build or test phase, potentially leading to remote code execution on the agent host.
- [COMMAND_EXECUTION]: The skill provides shell command templates, such as swiftlint lint --quiet <changed_files>, that are vulnerable to command injection if the agent populates the placeholder with unsanitized filenames containing shell metacharacters.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and review untrusted code. There are no explicit instructions or mechanisms provided to sanitize the content of these files, making the agent vulnerable to prompt injection attacks embedded within the code or comments of the reviewed repository.
Audit Metadata