phoenix-channels-essentials
Pass
Audited by Gen Agent Trust Hub on Aug 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions and code snippets follow standard security practices for the Phoenix framework, such as using token-based authentication in
connect/3and performing authorization injoin/3.\n- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges and documents the surface for ingesting untrusted data via Phoenix Channels, providing explicit security rules to mitigate associated risks.\n - Ingestion points: The skill provides examples of
handle_in/3functions that receive untrusted payloads from clients inSKILL.md.\n - Boundary markers: The skill includes a mandatory rule (Rule 7) requiring that client payloads are never passed raw to the LLM context and should be validated against a strict schema.\n
- Capability inventory: The provided logic includes broadcasting data to other users via
broadcast!and sending messages to the client viapush.\n - Sanitization: The skill demonstrates and recommends input sanitization techniques such as length limiting using
String.sliceand whitespace removal usingString.trim.
Audit Metadata