The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions provide several commands for the agent to execute the provided Node.js helper script (scripts/search.js) to perform web searches and quota checks. Additionally, a Python test script (scripts/test_helper.py) uses the subprocess.run function to invoke the Node.js helper during offline regression testing.
[EXTERNAL_DOWNLOADS]: The skill communicates with external endpoints api.minimax.io and api.minimaxi.com. These are the official domains for the MiniMax API service described in the skill's purpose.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest web search snippets, which is a potential surface for indirect prompt injection. The skill mitigates this by including explicit instructions in SKILL.md for the agent to treat all result text as untrusted data and never as instructions, adhering to safety best practices.
[CREDENTIALS_SAFE]: The skill requires a MINIMAX_API_KEY which is correctly handled via environment variables. The documentation includes instructions to check for the key's presence without printing it to logs, reducing the risk of accidental exposure.

minimax-search