atmos-automationlisted

# Atmos automation You are inside a **Terminal** or **Agent Chat** tab created by an Atmos automation run. The run stays `running` until you mark it complete. Process or TTY exit does **not** finish the job. Details: [`references/cli.md`](references/cli.md). ## When to use - The first prompt told you to read `~/.atmos/skills/.system/atmos-automation/SKILL.md`. - You are doing a saved Atmos automation (job + run artifacts). Do **not** use this skill to start automations from scratch, or for Headless Spawn runs (the runner writes `final.md` / `run.json` on process exit). ## Flow 1. Call `atmos automation paths --run <run_guid>` (never guess paths). 2. Read `instructions.md` at `instructions_path`. 3. Do the job in the returned `cwd`. 4. Write the final result to `result_path` (`final.md`). 5. Update `memory.md` only for a durable fact a later run would miss. 6. Mark the run finished: - Success: `atmos automation complete --run <run_guid>` - Failure: `atmos automation complete --run <run_guid> --failed --message "<short reason>"` Optional: `atmos automation status --run <run_guid>` to confirm. ## Rules - Never guess file paths. Always call `paths` first. - Never mark complete without a non-empty `final.md` unless `--failed`. - Do not treat process exit, TTY close, or chat idle as done. - Do not exec ad-hoc file writes outside the returned paths. - Do not invent other CLI verbs. Only `complete`, `status`, and `paths` (plus `run` if a human asked to start a job).