The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes the mmx CLI tool to perform various AI-driven tasks such as text completion, image generation, and speech synthesis. This involves executing subprocesses with parameters defined by user input.\n- [EXTERNAL_DOWNLOADS]: Instructions include the global installation of the mmx-cli package from the NPM registry, which is a standard procedure for using this tool.\n- [DATA_EXFILTRATION]: The skill documents the management of API keys and credentials, which are stored locally in the ~/.mmx/ directory. This is consistent with standard authentication practices for CLI-based service integrations.\n- [PROMPT_INJECTION]: The skill handles user-provided prompts and messages for media generation and chat, presenting a surface for indirect prompt injection. This is documented as a standard interaction model for AI CLI tools.\n
Ingestion points: CLI arguments such as --prompt, --message, and --q in SKILL.md\n
Boundary markers: None present\n
Capability inventory: Local command execution (mmx), network access (MiniMax API), and file system writes (--out, --download)\n
Sanitization: Not documented in the provided instructions

mmx-cli