The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read sensitive repository files, specifically .kamal/secrets, to gather context on existing credentials. Accessing files containing secrets is a standard part of Kamal's configuration workflow, though it creates a potential surface for data exposure if the agent's context is not properly contained.\n- [COMMAND_EXECUTION]: The skill documentation includes powerful CLI operations such as kamal accessory boot, kamal accessory reboot, and kamal accessory exec. These commands enable the agent to manage service lifecycles and execute arbitrary commands within remote containers, which are necessary functions for the documented tool.\n- [PROMPT_INJECTION]: The skill is designed to ingest and mirror external stack definitions, creating an indirect prompt injection surface.\n
Ingestion points: The agent reads config/deploy.yml, docker-compose.yml, and Helm charts to translate service definitions (SKILL.md).\n
Boundary markers: There are no explicit instructions or delimiters used to ensure the agent ignores natural language instructions that might be embedded in these external configuration files.\n
Capability inventory: The agent has access to container execution and server-side management via the kamal accessory subcommand suite (references/commands.md).\n
Sanitization: The skill does not define specific validation or sanitization protocols for the data extracted from external stack definitions before it is applied to the deployment configuration.

accessories