auth-state-sync

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [SAFE]: The skill instructions provide standard web development patterns for cross-tab state management using built-in browser APIs. The described methods for distributed locking and silent refresh are established industry practices.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements logic to ingest and act upon data from storage and message channels.
  • Ingestion points: Instructions in SKILL.md utilize window.addEventListener('storage') and BroadcastChannel.onmessage to receive external state updates.
  • Boundary markers: The logic restricts processing to specific expected keys (e.g., auth:logout) and message types (e.g., LOGOUT).
  • Capability inventory: The logic can trigger page redirections, update in-memory variables, and clear storage state.
  • Sanitization: The implementation uses JSON.parse() for deserialization and relies on specific property checks before execution.
  • [EXTERNAL_DOWNLOADS]: The skill references the jwt-decode library in its code examples for client-side token validation. This is a standard and well-known library for processing JWT metadata.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 10:43 PM
Security Audit — agent-trust-hub — auth-state-sync