| name | satelle |
| description | Route Satelle work to the narrow setup, use, or recover skill. Use when an agent is asked to configure, operate, troubleshoot, or explain Satelle but the correct Satelle workflow is not obvious. |
Satelle
Satelle is a remote computer-use bridge. Route to the smallest Satelle skill that can complete the job.
Route
- Use
satelle-setup when the task is to install, configure, onboard, or verify a Satelle host before prompt execution.
Completion criterion: the agent has a setup plan, knows whether mutation consent is needed, and knows the next verification command.
- Use
satelle-use when the task is to run, steer, inspect, stop, or follow an existing Satelle session.
Completion criterion: the agent has used the session contract instead of treating a one-shot command as disposable state.
- Use
satelle-recover when setup, run, steer, status, logs, MCP, or host commands fail or produce readiness blockers.
Completion criterion: the agent has a typed failure, evidence from diagnostics or logs, and a recovery path that preserves state.
- Stay in this router only for high-level explanation or when choosing among the other skills.
Completion criterion: the response names the specific skill to use next and why.
Discover the release-matched bundle with satelle skills list. Use satelle skills get <name> to read one guide or satelle skills path <name> when a client needs a local SKILL.md path.
Rules
- Prefer Satelle JSON output for agent decisions.
- Treat
session_id as the public handle. Do not depend on underlying Codex thread IDs.
- Do not handle raw provider secrets in chat, logs, facts, config snippets, or skill output.
- Do not use
--yes, YOLO mode, raw diagnostic exports, or recording modes unless the user explicitly asks for that risk-bearing behavior.
