/resume-mission

Resume the mission by launching (or reconnecting to) the autonomous runner.

Procedure

Step 1: Check State

Read .chariot/state.json and check current status.

• status: "completed" → "Mission already completed."
• status: "idle" or file missing → "No mission found. Run /chariot:set-mission first."
• status: "escalated" → Go to Step 2 (Escalation Handling).
• status: "pending_restart" / "paused" / "running" → Go to Step 3 (Runner Check).

Step 2: Escalation Handling

If status is "escalated":

1. Read escalations array in state.json for failure reasons
2. Read .chariot/worklog.md for failure history
3. Explain the failure to the user and offer choices:
   • Retry: Reset fail_count, retry the same task
   • Skip: Jump to the next task
   • Guided retry: User provides direction, then retry
4. After user's choice, update state.json accordingly (reset fail_count if retry, advance task if skip)
5. Proceed to Step 3.

Step 3: Check Runner Status

Check if a runner is already alive:

node -e "
const fs = require('fs');
const state = JSON.parse(fs.readFileSync('.chariot/state.json', 'utf-8'));
const pid = state.runner_pid;
if (pid) {
  try { process.kill(pid, 0); console.log('ALIVE:' + pid); }
  catch { console.log('DEAD:' + pid); }
} else {
  console.log('NONE');
}
"

If ALIVE:{pid} → Runner is already working. Show status:

Runner is already active (PID {pid}).
Log: tail -f .chariot/runner.log
Status: /chariot:status
Halt: /chariot:halt

No action needed. The runner continues autonomously.

STOP. Do not spawn a new runner.

If DEAD:{pid} → Previous runner crashed. Clean up and respawn (Step 4).

If NONE → No runner. Spawn one (Step 4).

Step 4: Spawn Runner

Use the dedicated launcher (bin/chariot-spawn.js). It handles the state update (status → running, reset fail counters, refresh plugin_root, clear stale runner_pid) AND strips parent Claude session env markers in one shot. Do NOT inline node -e blobs — zsh history expansion mangles ! characters and the spawn fails.

Read plugin_root from state.json (recorded by SessionStart hook), then run the launcher:

PLUGIN_ROOT=$(node -p "JSON.parse(require('fs').readFileSync('.chariot/state.json','utf-8')).plugin_root")
node "$PLUGIN_ROOT/bin/chariot-spawn.js"

The launcher prints {"pid": <runner_pid>, "current_task": "..."} on success. If it errors out (missing state.json, missing bin/chariot.js under plugin_root), the message goes to stderr and exit code is 1.

Step 5: Confirm Launch

Display to the user:

══════════════════════════════════════════════════════════
 CHARIOT RUNNER LAUNCHED
══════════════════════════════════════════════════════════

 Current task: Task {current_task}
 Progress:     {completed_count}/{total_tasks}
 Runner PID:   {pid}
 Log:          .chariot/runner.log

 The runner spawns fresh Claude sessions per task:
   → No context degradation (each task = clean slate)
   → Independent Inspector verification
   → Automatic commits on task completion

 Commands:
   /chariot:status     — Check progress
   /chariot:halt       — Pause execution
   tail -f .chariot/runner.log  — Live log

 You can close this session safely.
══════════════════════════════════════════════════════════

Step 6: Done

STOP. Do not start any implementation work. The runner handles everything independently. This Claude session's job is done.