backend-realtime

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFECREDENTIALS_UNSAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill instructs the agent to execute a command that reads from environment configuration files (.env*) to verify the presence of required Supabase keys.
  • Evidence: grep -oh "SUPABASE_[A-Z_]*" .env* 2>/dev/null | sort -u in SKILL.md (Line 49).
  • Context: The command is designed to only output variable names (keys) and not their values, mitigating the risk of secret exposure, but it still targets a sensitive file path.
  • [COMMAND_EXECUTION]: The skill uses shell commands to inspect the project structure, dependencies, and existing code patterns.
  • Evidence: Usage of cat, rg, and grep in SKILL.md to verify package.json content and search for existing WebSocket or Supabase implementations.
  • [INDIRECT_PROMPT_INJECTION]: The skill implements patterns for real-time data ingestion from Supabase Database Changes and Broadcast events, which could serve as an attack surface for indirect prompt injection if the ingested data is not sanitized before being processed by the agent or rendered in a way the agent observes.
  • Ingestion Points: payload.new in useRealtimeMessages and payload in useBroadcast (detected in references/patterns.md).
  • Boundary Markers: None present in the provided code snippets to delimit untrusted data.
  • Capability Inventory: The skill involves shell commands (cat, rg, grep) and state updates within a React environment.
  • Sanitization: The implementation examples do not include sanitization or validation of the incoming real-time payloads before updating the application state.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:26 PM