to-ios-probe
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the generation and execution of diagnostic code (probes) on target iOS devices or simulators.
- Evidence: The workflow utilizes sibling skills
to-ios-probe-scriptfor generation andto-ios-runfor execution of runtime operations. - This is the core functional purpose of the skill and relies on internal workspace tooling.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes external data from target applications which could potentially contain malicious instructions.
- Ingestion points: The skill reads Objective-C method chains, application state, and UI metadata (labels, identifiers) from external evidence files produced during probe execution.
- Boundary markers: The skill instructions do not explicitly define delimiters for separating untrusted application data from system instructions.
- Capability inventory: The skill can generate new probes and initiate executions based on its analysis of previous results.
- Sanitization: No explicit sanitization of UI labels or log messages is mentioned before they are processed by the agent context.
- This represents a standard attack surface for runtime analysis tools and is considered a low-severity inherent risk.
Audit Metadata