The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill designs a gateway function that ingests and processes untrusted message content intended for an LLM, creating a potential surface for indirect prompt injection if the downstream model is not properly guarded.
Ingestion points: The groqGateway function in references/implementation.md accepts a messages array representing user-provided conversation history.
Boundary markers: Absent; the messages are passed directly to the Groq SDK completion call.
Capability inventory: The implementation facilitates network communication with the Groq API and performs file logging of usage data.
Sanitization: Absent; the gateway is designed for authorization and budget enforcement rather than content moderation or sanitization of prompt data.
[COMMAND_EXECUTION]: The skill provides documentation for rotating API keys using standard CLI tools (aws and gcloud). These are presented as reference implementation steps for secure credential management rather than automated execution.
[CREDENTIALS_UNSAFE]: The skill uses gsk_... placeholders in code snippets to represent Groq API keys. These are clearly identifiable as synthetic placeholders for documentation purposes and do not represent leaked credentials. The instructions correctly advise storing real keys in professional secret management solutions.

groq-enterprise-rbac