axiom-ios-build
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill operates as a navigation and documentation layer for iOS development, providing structured logic for diagnosing build failures, simulator issues, and dependency conflicts. It functions primarily through instructional routing without executing unauthorized scripts or network requests.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes workflows that ingest untrusted external data, specifically iOS crash reports (.ips/.crash files) and TestFlight feedback, which can serve as a vector for malicious instructions targeting the AI agent during the analysis phase.
- Ingestion points: User-provided crash log text and references to local diagnostic files (e.g.,
~/Library/Logs/DiagnosticReports/) described in the example invocations withinSKILL.md. - Boundary markers: The instructions lack explicit delimiters or warnings to ignore potentially adversarial instructions embedded within the logs or feedback text.
- Capability inventory: The router directs users to autonomous agents (e.g.,
build-fixer,crash-analyzer) that possess capabilities to perform environment diagnostics and parse complex payloads. - Sanitization: No sanitization or validation logic is defined in the skill to filter or escape content within the ingested diagnostic reports before they are processed by the LLM.
Audit Metadata