ainb-fleetdaemonlisted

Long-running watcher that scans every claude session every 5s and auto-sends `continue` to any session whose recent tmux pane buffer matches a known API-error regex (rate_limited, overloaded_error, internal_server_error, request_timeout, socket_hang_up, fetch_failed, ECONNRESET). Use this when you want unattended recovery from transient API failures across the fleet.
stevengonsalvez/agents-in-a-box · ★ 22 · AI & Automation · score 78

Install: claude install-skill stevengonsalvez/agents-in-a-box

# ainb fleet:daemon > **DEPRECATED. Use [`/ainb-fleet:atc`](../atc/SKILL.md) in `lite` mode.** > `ainb fleet atc mode <name> --set lite` gives you this exact loop — the same > 5-second scan, the same auto-`continue` on the same known transient errors — > owned by the ATC supervisor, inside its per-session retry cap and its safety > ledger, and structurally unable to run beside the LLM heartbeat. > > This standalone daemon REFUSES to start while ATC owns the fleet in EITHER > mode, because both send the same auto-`continue` to the same pane and each > de-dups only inside its own process, so the pane gets it twice. Pass > `--force-race` to run both anyway. > > The difference that matters is the **retry cap**: ATC gives up on a session > after a bounded number of attempts and escalates to you. This daemon has no > cap, so a session that keeps failing is retried forever, and running it beside > ATC makes ATC's cap meaningless. > > Kept for unmanaged one-off recovery, not removed. > **Superseded by [`/ainb-fleet:atc`](../atc/SKILL.md) for managed fleets.** ATC > absorbs this daemon's job — its ERR playbook does the same auto-`continue`, but > **with a per-session retry cap + escalate-on-exhaustion** that this daemon > lacks (see the sharp edge below). Prefer `ainb fleet atc setup <name>` for > unattended supervision. The daemon stays for unmanaged one-off recovery; do > **not** run it against sessions an ATC instance already manages — they race. Background watcher. Reads each s