
This handles the full lifecycle of agents on the TerminalUse platform: scaffolding with `tu init`, deploying to prod or preview environments, and running tasks against them. The workflow is straightforward: you initialize an agent directory, deploy it, then create tasks with either an existing filesystem or a project ID that auto-provisions one. The skill emphasizes context awareness, checking if you're in an agent directory before running commands and preferring explicit flags over implicit defaults. One thing worth noting is the clear anti-patterns section, like warning you not to mess with Dockerfile entrypoints or expect instant secret propagation. It's opinionated about the happy path but gives you escape hatches when you need them.
npx -y skills add terminaluse/agent-skills --skill terminaluse --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.