internet-identity-notifications

Pass

Audited by Gen Agent Trust Hub on Nov 7, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [SAFE]: The skill provides legitimate technical documentation for Internet Identity notifications. It includes several security-focused instructions, such as validating that notification URLs match the sending origin and ensuring that consent is handled via secure user gestures.
  • [EXTERNAL_DOWNLOADS]: The skill references external dependencies including the @icp-sdk/auth NPM package and the identity-notifications library (available via crates.io for Rust and Mops for Motoko). These are standard libraries within the Internet Computer Protocol (ICP) ecosystem and are maintained by the platform's developers.
  • [INDIRECT_PROMPT_INJECTION]: The skill describes an interface for sending notifications that ingests dynamic content, creating a potential surface for indirect prompt injection if user-provided data is not sanitized by the implementing canister.
  • Ingestion points: The title, body, and url fields passed to the send() function in the Rust and Motoko examples.
  • Boundary markers: The skill instructions do not specify content-level delimiters for the message body.
  • Capability inventory: The capability is limited to sending push notifications to signed-in users' browsers.
  • Sanitization: The documentation explicitly states that Internet Identity validates the url field against the sending origin or alternative origins to prevent malicious redirection.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 7, 2026, 02:08 PM