automation-reference

Automate CLI and API tasks by applying an automation-first principle where agents execute any tool with a CLI or API, reserving human actions solely for non-automatable verification checks. This handles environment variables, dev-server lifecycles, and installation flows.

An automation-first principle routes automatable development, deployment, and maintenance work to agents rather than humans. It ensures agents perform CLI or API operations directly, minimizing manual effort across service provisioning and configuration management workflows.

Manage dev-server lifecycles and environment variables by instructing agents to handle these CLI or API operations directly. Agents execute the automatable steps, applying safety constraints and verification checkpoints to ensure governed, safe operations.

Automation-first agent routing works for configuration management by applying to CLI or API-based services across projects. It supports deployment pipelines and service provisioning, ensuring agents execute automatable tasks while humans perform only verification.

Limitations include the need for explicit safety constraints to prevent prompt injection or data exfiltration. Agents require YAML frontmatter defining name, description, type, and agents in SKILL.md, and humans must still perform non-automatable verification checkpoints.

You need YAML frontmatter in SKILL.md to set up automation workflows, defining name, description, type, agents, and sections. Optional resources directories for scripts, references, and assets can also be included to support the workflow.