Aigent.ly — Vulnerability Prevention for AI-Generated Code

Rule body

AI Summary

AI Security Summary

iOS system-level components like RTKit present a privileged attack surface reachable from app code and native extensions. The recurring mistake is failing to validate and sandbox memory operations at trust boundaries between user-space and firmware-level subsystems.

Firmware-Level Memory Corruption via RTKit Trust Boundary

Covers: CVE-2024-23296 Risk: ACTIVELY EXPLOITED memory corruption in RTKit allows attackers to escape app sandboxes and execute arbitrary code at the firmware level.

NEVER pass unvalidated or attacker-influenced data directly to low-level system or firmware interfaces exposed through iOS frameworks.
ALWAYS treat any buffer or memory region shared between user-space and system co-processors as an untrusted, adversarial boundary.
NEVER assume that system-level components beneath the OS kernel enforce their own input safety — validate length, type, and bounds before any cross-layer call.
ALWAYS use the highest available abstraction layer (e.g., AVFoundation, CoreImage) instead of direct low-level system API calls that interact with firmware subsystems.
NEVER allow user-supplied integers or offsets to directly index into memory buffers passed to kernel or firmware extensions without explicit bounds checking.

Cross-cutting patterns (all iOS / Swift projects)

ALWAYS enable all compiler-enforced memory safety features (e.g., Address Sanitizer, stack canaries) during development and testing to surface corruption bugs before production.
NEVER share mutable memory buffers across privilege boundaries without explicit ownership transfer and lifecycle control.
ALWAYS audit any Swift code that bridges to Objective-C or C APIs for unsafe pointer arithmetic, force-unwrapped optionals on memory addresses, and unchecked casts.

View all CVE patterns

# iOS / Swift security patterns

Safe, always-on code patterns that reduce security risk without changing your dependency graph.

These rules use ONLY pattern-level verbs (ALWAYS/NEVER). They MUST NOT prescribe dependency upgrades.

### CVE-2024-23296 — safe patterns

- NEVER construct RTKit messages from untrusted or user-controlled input without strict validation.
- ALWAYS use type-safe serialization when building messages for RTKit communication.
- ALWAYS validate message boundaries and structure before passing to RTKit interfaces.
- NEVER assume RTKit will reject malformed message formats—validate before sending.

# iOS / Swift security patterns Safe, always-on code patterns that reduce security risk without changing your dependency graph. These rules use ONLY pattern-level verbs (ALWAYS/NEVER). They MUST NOT prescribe dependency upgrades. ### CVE-2024-23296 — safe patterns - NEVER construct RTKit messages from untrusted or user-controlled input without strict validation. - ALWAYS use type-safe serialization when building messages for RTKit communication. - ALWAYS validate message boundaries and structure before passing to RTKit interfaces. - NEVER assume RTKit will reject malformed message formats—validate before sending.

AI Security Summary

Firmware-Level Memory Corruption via RTKit Trust Boundary

Covers: CVE-2024-23296 Risk: ACTIVELY EXPLOITED memory corruption in RTKit allows attackers to escape app sandboxes and execute arbitrary code at the firmware level.

NEVER pass unvalidated or attacker-influenced data directly to low-level system or firmware interfaces exposed through iOS frameworks.

ALWAYS treat any buffer or memory region shared between user-space and system co-processors as an untrusted, adversarial boundary.

NEVER assume that system-level components beneath the OS kernel enforce their own input safety — validate length, type, and bounds before any cross-layer call.

ALWAYS use the highest available abstraction layer (e.g., AVFoundation, CoreImage) instead of direct low-level system API calls that interact with firmware subsystems.

NEVER allow user-supplied integers or offsets to directly index into memory buffers passed to kernel or firmware extensions without explicit bounds checking.

Cross-cutting patterns (all iOS / Swift projects)

ALWAYS enable all compiler-enforced memory safety features (e.g., Address Sanitizer, stack canaries) during development and testing to surface corruption bugs before production.

NEVER share mutable memory buffers across privilege boundaries without explicit ownership transfer and lifecycle control.

ALWAYS audit any Swift code that bridges to Objective-C or C APIs for unsafe pointer arithmetic, force-unwrapped optionals on memory addresses, and unchecked casts.

Protects against

Rule body

AI Security Summary

Firmware-Level Memory Corruption via RTKit Trust Boundary

AI Security Summary

Firmware-Level Memory Corruption via RTKit Trust Boundary