How It Works
How It Works
Section titled “How It Works”The SuperDev workflow starts from a goal, not from a panel.
Tell the agent:
Create this repository as a SuperDev project with the services it needs.
The agent should then:
- Inspect scripts, manifests, Docker files, README files, and existing deployment notes.
- Probe the project with SuperDev tools.
- Propose project, service, deployment, and pipeline configuration.
- Preview config changes before applying them.
- Wait for you in the desktop app when a write operation needs approval.
- Start or attach to services, read logs, open browser sessions, or collect pipeline evidence.
- Report what it verified and which assumptions remain open.
What Agent-first Means
Section titled “What Agent-first Means”Agent-first does not ask you to learn a manual procedure first. You state the goal and boundaries, then review the agent’s project model, config diff, approval prompt, and runtime evidence.
You still own boundaries: secrets, destructive actions, production targets, and final approval. The agent owns the boring runtime work.
Why The Skill Matters
Section titled “Why The Skill Matters”The installed SuperDev skill tells the agent the operating discipline: start with list_services or get_runtime_snapshot, use tail_logs instead of asking for pasted terminal output, use browser tools for local frontend verification, and collect evidence before claiming a fix.