kmp-ktor-realtime

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill implements a structured approach to real-time communication using standard Ktor 3 and Kotlin Coroutines. No malicious code or unauthorized access patterns were identified.
  • [SAFE]: Resource management is handled via bounded channels (inbound and outbound), effectively preventing memory exhaustion or Denial of Service (DoS) from slow or malicious clients.
  • [SAFE]: Thread safety is achieved through a single-threaded authoritative loop using a limited parallelism dispatcher. This design avoids common concurrency vulnerabilities and race conditions.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external WebSocket clients, creating a potential surface for indirect prompt injection if the output is processed by an AI agent.
  • Ingestion points: Incoming WebSocket frames are read and decoded in the webSocket route within SKILL.md.
  • Boundary markers: The code uses a codec (JsonMessageCodec) and typed routing (MessageRouter) to validate incoming data types and delimit messages.
  • Capability inventory: The skill does not perform any dangerous operations such as shell execution, file system modifications, or unauthorized network requests.
  • Sanitization: Messages are decoded into typed objects, preventing the execution of arbitrary strings as commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 03:04 PM