CI and AI agents

cloudsesame is safe to run where no one can answer a prompt. Without a terminal it never hangs: it fails with exit code 4 and prints the exact commands to run, or continues with --skip.

Without a terminal

There's no terminal in CI, in coding agents, with piped input, or with --no-input or --json. A prompt then becomes what noTerminal says: "fail" (the default) or "skip".

✖ Not ready: aws prod
  No terminal to prompt in. Run `aws sso login --profile prod`, or rerun with --skip to continue without it.

• --skip (or noTerminal: "skip") continues without the provider, and continues offline in predev.
• --strict makes skipped providers count as failures.
• The open-sesame animation never plays without a terminal, in CI (CI set) or with NO_COLOR.

See Recipes: CI pipelines for choosing between them.

JSON output

cloudsesame check --json never prompts and prints the result as JSON:

{
  "ok": false,
  "code": 4,
  "online": true,
  "loggedIn": false,
  "results": [
    { "provider": "gcloud", "label": "adc", "state": "ok" },
    {
      "provider": "gcloud",
      "label": "project",
      "state": "mismatch",
      "identity": "acme-sandbox",
      "message": "expected project acme-dev",
      "fix": ["gcloud", "config", "set", "project", "acme-dev"]
    }
  ]
}

Field	Description
ok	true when code is 0.
code	The exit code.
online	Result of the connectivity check, or null with onlineCheck: false.
loggedIn	Whether a login ran during this run.
results[]	One entry per check: provider, label, state (ok · unauthenticated · mismatch · denied · missing · error), and when known identity, message, fix (the command as an argv array) and skipped.

--json works with auth and predev too, and implies --no-input. See Providers for what each label checks.

Agent Skill

cloudsesame ships an Agent Skill that teaches coding agents to set it up in a project and to handle its failures: ask you to run logins rather than attempting them, and use --skip when the task doesn't need the cloud.

cloudsesame init offers to install it. Or install it with the skills CLI, which supports Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot and many more:

bunx skills add NexiHealth/cloudsesame          # or: npx skills add NexiHealth/cloudsesame
bunx skills add NexiHealth/cloudsesame -g       # for all your projects
bunx skills update                              # pull the latest version later

Without the skills CLI, copy skills/cloudsesame/ (also in the npm package, under node_modules/cloudsesame/skills/) into your agent's skills folder, e.g. .claude/skills/.

Why agents shouldn't log in

Logins open a browser and need you. The skill tells agents never to pass --yes or --unauthenticated login, and to show you the printed commands instead.