foundation-models-auditor

Scans Foundation Models (Apple Intelligence) code for the 10 most critical violations — missing availability checks, main thread blocking, manual JSON parsing, missing error handling, and session lifecycle issues.

How to Use This Agent

Natural language (automatic triggering):

• "Can you check my Foundation Models code for issues?"
• "Review my @Generable structs for correctness"
• "Audit my Apple Intelligence integration"
• "My LanguageModelSession keeps crashing"

Explicit command:

/axiom:audit foundation-models

What It Does

1. No Availability Check (CRITICAL) — Crash on devices without Apple Intelligence
2. Synchronous respond() on Main Thread (CRITICAL) — UI freeze and watchdog kill
3. Manual JSON Parsing (CRITICAL) — Use @Generable instead of JSONDecoder
4. Missing exceededContextWindowSize Catch (HIGH) — Needs conversation trimming
5. Missing guardrailViolation Catch (HIGH) — Needs user-facing safety messaging
6. Session Created in Button Handler (HIGH) — Wasteful recreation on every tap
7. No Streaming for Long Generations (MEDIUM) — Unresponsive UI during inference
8. Missing @Guide Annotations (MEDIUM) — Unconstrained numeric/collection output
9. Nested Type Without @Generable (MEDIUM) — Compilation or runtime failure
10. No Fallback UI When Unavailable (LOW) — Broken UI on unsupported devices

Related

• foundation-models — On-device AI implementation patterns (iOS 26+)
• foundation-models-diag — Foundation Models troubleshooting
• foundation-models-ref — Complete API reference with WWDC examples