shareplay-activities

Pass

Audited by Gen Agent Trust Hub on Sep 11, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implementation patterns for GroupSessionMessenger and GroupSessionJournal define a protocol for ingesting untrusted data from external participants, which represents an indirect prompt injection surface.\n
  • Ingestion points: Untrusted data enters the application context through the messenger.messages(of:) stream in DrawingManager, SharePlayManager, and QuizManager (in references/shareplay-patterns.md).\n
  • Boundary markers: The code examples do not implement boundary markers or instructions to disregard potential commands embedded in shared messages.\n
  • Capability inventory: The documented capabilities are limited to state synchronization and UI updates; no file-writing or command-execution tools are present in the skill.\n
  • Sanitization: The patterns do not include logic for sanitizing or validating the content of the Codable messages before they are appended to local state.\n- [EXTERNAL_DOWNLOADS]: The skill contains multiple reference links to a non-standard domain (sosumi.ai) for documentation, rather than official vendor resources.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 11, 2026, 11:13 AM
Security Audit — agent-trust-hub — shareplay-activities