The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to run cat .env* | grep -i SUPABASE to verify configuration. This command reads sensitive environment files (such as .env, .env.local, or .env.production) that typically contain private credentials like API keys and database secrets, exposing them to the agent's context.
[COMMAND_EXECUTION]: Shell commands including cat, grep, and rg are utilized to inspect project dependencies, file contents, and environment variables.
[PROMPT_INJECTION]: The implementation of real-time data listeners creates a surface for indirect prompt injection.
Ingestion points: Untrusted data enters the agent's context through postgres_changes payloads and broadcast event payloads in Supabase subscriptions as shown in the provided React hooks.
Boundary markers: There are no explicit boundary markers or instructions to ignore potential commands embedded within the real-time data stream.
Capability inventory: The skill manages persistent network connections and processes database events that directly influence the application's state and data flow.
Sanitization: The code examples lack validation or sanitization logic to filter external payload content before it is rendered or processed.

backend-realtime