Help us improve
Share bugs, ideas, or general feedback.
From autonomous-dev
Resume autonomous execution from checkpoint after session interrupt or token limit
npx claudepluginhub adiomas/claude-code-adiomas-plugins --plugin autonomous-devHow this command is triggered — by the user, by Claude, or both
Slash command
/autonomous-dev:auto-continueThis command is limited to the following tools:
The summary Claude sees in its command listing — used to decide when to auto-load this command
# /auto-continue - Resume Autonomous Session Resume an interrupted autonomous development session from the last checkpoint. ## When to Use - After session was interrupted (network, timeout, manual stop) - After token limit triggered graceful handoff - When continuing work in a new session - After `/auto-status` shows incomplete session ## Resume Process ### Step 1: Detect Previous Session Read the state machine to understand where we left off: - Current state (DETECT, CLASSIFY, PLAN, EXECUTE, INTEGRATE, REVIEW) - Work type (FRONTEND, BACKEND, etc.) - Completed phases - Checkpoint fi...
/buidl-resumeResumes an interrupted loop session from its last checkpoint. Reports session status, phase, cycle, worktree path, branch, and PR details, then continues execution.
Share bugs, ideas, or general feedback.
Resume an interrupted autonomous development session from the last checkpoint.
/auto-status shows incomplete session# Check for existing state machine
if [[ -f ".claude/auto-state-machine.yaml" ]]; then
echo "Previous session found"
RESUME_MODE=true
else
echo "No previous session to resume"
exit 0
fi
Read the state machine to understand where we left off:
Read .claude/auto-memory/ files to reconstruct context:
CRITICAL: Do NOT re-run completed phases. Trust the checkpoints.
Based on current_state in state machine:
| Current State | Resume Action |
|---|---|
| DETECT | Re-run detection (rare - usually complete) |
| CLASSIFY | Read auto-context.yaml, continue to PLAN |
| PLAN | Read existing plan, get user approval, continue to EXECUTE |
| EXECUTE | Check auto-progress.yaml for pending tasks, continue execution |
| INTEGRATE | Continue merging branches |
| REVIEW | Run fresh verification |
| RESEARCH | Continue from current research phase |
For EXECUTE phase resume:
# Read .claude/auto-progress.yaml to find:
tasks:
task-1:
status: done # ← Skip
task-2:
status: done # ← Skip
task-3:
status: in_progress # ← RESUME HERE
branch: auto/task-3
iterations: 2
task-4:
status: pending # ← Do after task-3
/tmp/auto-worktrees/task-{id}/Before continuing, validate:
# Verify plan file exists
ls .claude/plans/auto-*.md
# Verify project profile exists
ls .claude/project-profile.yaml
# Verify auto-context exists
ls .claude/auto-context.yaml
If any are missing, notify user and ask how to proceed.
On successful resume:
=== RESUME SESSION ===
Session ID: auto-20240115-103000
Resuming from: EXECUTE phase
Work Type: FRONTEND
Completed: DETECT, CLASSIFY, PLAN
Pending Tasks:
- task-3: Create Auth API (in_progress, iteration 2)
- task-4: Add Tests (pending)
Mandatory Skills for EXECUTE:
- superpowers:test-driven-development
- frontend-design
Continuing execution...
=== END RESUME ===
No previous autonomous session found.
Use /auto to start a new session.
State machine file is corrupted or incomplete.
Options:
1. Delete .claude/auto-state-machine.yaml and start fresh with /auto
2. Manually fix the state file
3. Check .claude/auto-memory/ for recoverable context
Plan file not found at .claude/plans/auto-*.md
Cannot resume EXECUTE phase without plan.
Options:
1. Re-run /auto-plan to create new plan
2. Check if plan was moved to docs/plans/
This command:
.claude/auto-state-machine.yaml for current state.claude/auto-memory/*.md for contextUser: The session was interrupted. Let me continue.
> /auto-continue
Claude: Resuming autonomous session...
Session ID: auto-20240115-103000
Current State: EXECUTE
Work Type: FRONTEND
Reading context from checkpoints...
- Phase 1 (DETECT): Complete - Next.js project detected
- Phase 2 (CLASSIFY): Complete - FRONTEND with DESIGN_QUALITY focus
- Phase 3 (PLAN): Complete - 5 tasks planned
Pending tasks:
- task-3: Create UserProfile component (in_progress)
- task-4: Add unit tests (pending)
- task-5: Integration tests (pending)
Invoking mandatory skills for EXECUTE phase:
- superpowers:test-driven-development
- frontend-design
Continuing with task-3...