The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents commands such as uc service exec SERVICE and uc exec web /bin/sh -c "env" that allow the agent to perform arbitrary shell execution within managed containers. This is a primary intended function for cluster administration.
[INDIRECT_PROMPT_INJECTION]: The skill enables an attack surface where instructions in external data could influence the agent.
Ingestion points: The agent is instructed to use uc service logs to inspect runtime output and uc deploy which processes local compose.yaml and .caddyfile configurations.
Boundary markers: The instructions do not define clear delimiters or warnings to the agent to disregard instructions found within the logs or configuration files it is tasked to inspect.
Capability inventory: The agent has access to powerful administrative commands including container shell access (uc exec), machine bootstrapping (uc machine init), and service lifecycle management (uc deploy, uc service rm) as defined in SKILL.md.
Sanitization: There is no mention of filtering, escaping, or validating the content of logs or configuration templates before the agent processes them.

uncloud