permissionkit
Pass
Audited by Gen Agent Trust Hub on Sep 11, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONMETADATA_POISONING
Full Analysis
- [METADATA_POISONING]: The skill provides detailed instructions for a non-existent framework and futuristic software versions. This deceptive content can lead to the generation of non-functional code and incorrect security expectations regarding child safety features.
- [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection where untrusted external data is incorporated into sensitive system prompts.
- Ingestion points: Untrusted strings are ingested via CommunicationHandle values, PersonInformation name components, and SignificantAppUpdateTopic descriptions in SKILL.md and references/permissionkit-patterns.md.
- Boundary markers: There are no delimiters or boundary markers to isolate user-provided strings within the permission request flow.
- Capability inventory: The skill utilizes AskCenter.shared.ask to present information to users and AskCenter.shared.responses to monitor and react to user choices.
- Sanitization: The provided patterns lack input sanitization or validation logic for the data processed by the permission system.
Audit Metadata