The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides curl and export command examples to the user for configuration and connectivity testing. These are standard developer setup instructions and are not executed automatically by the agent without user intent.
[DATA_EXFILTRATION]: The skill communicates with official ZenMux endpoints (zenmux.ai) for API services and documentation. These interactions are core to the skill's purpose as a vendor setup guide and do not involve unauthorized data transfer.
[CREDENTIALS_UNSAFE]: Configuration steps involve handling API keys, but the skill correctly uses placeholders (e.g., <your-zenmux-api-key>) and instructs users to store them securely in environment variables rather than hardcoding them.
[INDIRECT_PROMPT_INJECTION]: The skill reads from local documentation files in .context/references/. While this constitutes an ingestion point for untrusted data if the references were to be poisoned, in this context, it represents reading vendor-provided documentation and is a standard platform feature with no evidence of malicious intent.

zenmux-setup