The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on executing the muapi CLI through several shell scripts (run-workflow.sh, generate-workflow.sh, etc.) to perform operations. The scripts use proper shell quoting and array handling to prevent command injection.
[EXTERNAL_DOWNLOADS]: The documentation instructs users to install or upgrade the muapi-cli from a local workspace path. It also communicates with the muapi.ai API to fetch workflow definitions and results, which are legitimate operations for this vendor-specific tool.
[PROMPT_INJECTION]: The skill processes workflow metadata (names and descriptions) retrieved from the remote API. This is an indirect prompt injection surface; however, the skill includes explicit 'CRITICAL RULES' for the agent to avoid guessing inputs and to require user confirmation, which effectively mitigates the risk of autonomous malicious behavior.

muapi-workflow