The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous examples of executing kamal CLI commands to remove containers, images, and data. It specifically mentions the -y flag, which allows these destructive operations to proceed without interactive user confirmation.
[DATA_EXFILTRATION]: The instructions direct the agent to read .kamal/secrets and config/deploy.yml to confirm environment details. These files are standard locations for deployment credentials (e.g., Docker registry passwords). While necessary for the deployment tool to function, this practice exposes sensitive data to the agent's execution context.
[EXTERNAL_DOWNLOADS]: The skill and its reference files contain links to official documentation at kamal-deploy.org. These are legitimate references to the tool's official resources and do not represent a security risk.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface as it is designed to ingest and act upon data from local configuration files.
Ingestion points: SKILL.md identifies config/deploy.yml and .kamal/secrets as primary data sources.
Boundary markers: No delimiters or safety instructions are provided to the agent regarding the content of these files.
Capability inventory: The agent is granted the ability to execute kamal commands which can alter or delete server infrastructure.
Sanitization: There is no evidence of content validation or sanitization for the ingested configuration files.

remove