The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs destructive operations including cf undeploy (which can delete cloud service instances and keys) and rm -rf to remove local directories. These operations are the primary purpose of the skill and are mitigated by explicit instructions requiring separate user confirmation for every destructive step.
[INDIRECT_PROMPT_INJECTION]: The skill reads from mta.yaml to extract the MTA ID. This is an external data ingestion point. If a project contains a malicious mta.yaml file, it could theoretically attempt to influence the agent's behavior or inject characters into shell commands. However, the skill's structure and the requirement for user confirmation of the extracted ID mitigate this risk.

cleanup