audioaccessorykit
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [METADATA_POISONING]: The skill claims to target 'iOS 26+', which is a non-existent version of the operating system. This is likely a hallucination or placeholder that could mislead developers regarding the framework's actual availability and requirements.
- [INDIRECT_PROMPT_INJECTION]: The skill describes patterns for ingesting data from external Bluetooth accessory sensors to update system configuration, creating a potential attack surface. 1. Ingestion points: The skill ingests Bluetooth identifiers and raw hardware sensor data via ASAccessory and PlacementMonitor in 'references/audioaccessorykit-patterns.md'. 2. Boundary markers: The instructions lack boundary markers or explicit instructions to the agent to treat hardware-provided data as untrusted. 3. Capability inventory: The skill demonstrates capabilities to register devices and update system-level audio routing via AccessoryControlDevice in 'SKILL.md' and 'references/audioaccessorykit-patterns.md'. 4. Sanitization: The patterns map sensor states to enums but lack validation for Data identifiers used as audio source markers.
- [DATA_EXFILTRATION]: The documentation references an untrusted external domain (sosumi.ai) for technical guides. These static links point to a non-whitelisted and non-official source.
Audit Metadata