/bmad-loop

/bmad-loop - Run BMAD Epics in Autonomous Loop

Process BMAD epics continuously using the autonomous loop from claude-auto-agents plugin.

Usage

/bmad-loop [epic-pattern]
/bmad-loop "7A 8A 10B"
/bmad-loop "sprint-12.*"

Requirements

This command requires the claude-auto-agents plugin to be installed for autonomous loop functionality.

Behavior

1. Scans for BMAD epic files matching the pattern
2. Adds each epic as a work item to work/queue.md
3. Starts the autonomous loop
4. For each epic:
   • Creates feature branch
   • Implements stories with TDD
   • Runs code review
   • Creates PR
   • Monitors CI and fixes issues
   • Merges when approved
5. Continues to next epic until queue empty

Epic Discovery

Epics are found in:

• _bmad-output/epics/
• _bmad-output/stories/epic-*/
• docs/epics/

Work Queue Format

Each epic is added to the queue as:

- [ ] **[EPIC-{id}]** Implement epic {id}: {title}
  - Priority: high
  - Agent: bmad-orchestrator

Monitoring

# Check loop status
/status

# View queue
/queue list

# Stop the loop
/stop

Example

# Process all epics in sprint 12
/bmad-loop "sprint-12.*"

# Process specific epics
/bmad-loop "7A 8A"

# Process all available epics
/bmad-loop

STATUS Signal

STATUS: COMPLETE | BLOCKED | WAITING | ERROR
SUMMARY: Brief description of epic processing status
FILES: comma-separated list of changed files
NEXT: Next epic to process or completion message
BLOCKER: Reason if blocked (e.g., CI failure, review required)

Autonomous Execution

When invoked, this command will:

1. Discover Epics: Search for epic files matching the pattern
2. Queue Epics: Add each epic to work/queue.md with proper format
3. Start Loop: Activate the autonomous loop via loop-control
4. Orchestrate: The bmad-orchestrator agent handles each epic

Step 1: Discover and Queue Epics

First, find epic files and add them to the work queue:

# Find epic files
PATTERN="${ARGUMENTS:-*}"
EPIC_DIR="_bmad-output/epics"

# Initialize work directory if needed
mkdir -p work

# Find matching epics and queue them
for epic_file in "$EPIC_DIR"/*.md; do
    if [[ -f "$epic_file" ]]; then
        epic_id=$(basename "$epic_file" .md)
        epic_title=$(head -1 "$epic_file" | sed 's/^#* *//')

        # Add to queue if matches pattern
        if [[ "$epic_id" == *"$PATTERN"* ]] || [[ "$PATTERN" == "*" ]]; then
            echo "- [ ] **[EPIC-$epic_id]** Implement: $epic_title" >> work/queue.md
            echo "  - Priority: high" >> work/queue.md
            echo "  - Agent: bmad-orchestrator" >> work/queue.md
        fi
    fi
done

Step 2: Activate the Loop

The loop will be started automatically. Epic pattern: $ARGUMENTS

After discovering epics, the orchestrator will:

1. Read the epic file to understand requirements
2. Create a feature branch
3. Implement each story with TDD
4. Request code review
5. Create PR and monitor CI
6. Fix any issues
7. Merge when approved
8. Emit STATUS: COMPLETE and move to next epic

Agent Instructions

You are the bmad-orchestrator. For each epic in the queue:

1. Read the epic file from _bmad-output/epics/ or _bmad-output/stories/
2. Create feature branch: git checkout -b feature/epic-{id}
3. For each story in the epic:
   • Write tests first (TDD)
   • Implement the feature
   • Ensure tests pass
4. Request review: Use the reviewer agent or /spawn reviewer
5. Create PR: gh pr create --title "Epic {id}: {title}"
6. Monitor CI: Wait for checks, fix if needed
7. Merge: When approved and CI passes
8. Emit STATUS:

   STATUS: COMPLETE
   SUMMARY: Implemented epic {id}: {title}
   FILES: list of changed files
   NEXT: Process next epic or "All epics complete"
   

If blocked:

STATUS: BLOCKED
SUMMARY: Cannot proceed with epic {id}
BLOCKER: Specific reason (e.g., "CI failing on test X", "Needs architecture decision")