axiom-audit-foundation-models
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is a purely instructional auditor for Swift projects using Apple's Foundation Models framework. It provides specific patterns for detecting crashes, thread blocking, and security vulnerabilities like prompt injection within the codebase being audited.
- [INDIRECT_PROMPT_INJECTION]: As an auditing tool, the skill is designed to process untrusted data (source code). This inherent attack surface is mitigated by the skill's highly structured instructions and focus on static analysis (Grep/Glob/Read). The skill even includes explicit checks for sanitization and prompt injection within the target code it audits, reflecting a security-conscious design for its primary purpose.
- [PROMPT_INJECTION]: The static analysis hint regarding 'ignore instructions' is a false positive. The phrase occurs within the skill's Phase 3 reasoning section as a description of an attack vector (prompt injection) that the auditor should look for in user-supplied strings within the target application code, rather than an attempt to subvert the agent's own constraints.
Audit Metadata