shazamkit
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEMETADATA_POISONINGINDIRECT_PROMPT_INJECTION
Full Analysis
- [METADATA_POISONING]: The reference section contains links to documentation on an unofficial domain (sosumi.ai) instead of the official Apple Developer portal. While not inherently malicious, these links represent an unverifiable third-party source for technical documentation.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface area for indirect prompt injection through its ingestion of external audio and catalog data.\n
- Ingestion points: Audio PCM buffers are ingested from the microphone via SHManagedSession and AVAudioEngine (SKILL.md). Additionally, SHCustomCatalog loads data from external .shazamcatalog files (SKILL.md, references/shazamkit-patterns.md).\n
- Boundary markers: Absent. The instructions do not define delimiters or provide guidance on handling potentially adversarial audio signals or malformed catalog files.\n
- Capability inventory: The skill facilitates audio matching and metadata extraction (SKILL.md, references/shazamkit-patterns.md). It does not include high-risk capabilities such as arbitrary command execution or network exfiltration based on matching results.\n
- Sanitization: Input processing is deferred to the system-level ShazamKit framework, and the skill does not implement additional validation or filtering for the audio stream or catalog content.
Audit Metadata