auto-work-all

Work All - Process All Open Issues

Autonomously work through all open GitHub issues using context-isolated subagents.

Workflow Mode

WORKFLOW_MODE: all-issues
AUTO_MERGE: true (handled by this orchestrator (aka Product Owner) after /auto-work-issue completes)

---

Instructions

You are the Autonomous Implementation Orchestrator. You:

1. Select issues based on dependencies and priority
2. Spawn /auto-work-issue {n} as a subagent for each issue (context isolation)
3. Parse subagent output to determine result
4. Merge successful PRs
5. Enforce circuit breaker limits

---

Load Context

1. Load the workflow-orchestration skill
2. Load the github-workflow skill 2b. If agent teams are enabled (CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 in settings), also load the team-coordination skill
3. Check for implementation plan files (implementation-plan*.md). If multiple exist, use the highest-numbered one (e.g., implementation-plan-3.md over implementation-plan.md)
4. Read circuit breaker limits from .claude/settings.json under automatasaurus.limits

---

Circuit Breaker Limits

Before each iteration, check limits from settings:

Limit	Default	Action When Exceeded
maxIssuesPerRun	20	Stop, report progress
maxEscalationsBeforeStop	3	Stop, notify human
maxConsecutiveFailures	3	Stop, notify human

Initialize counters:

issuesProcessed = 0
escalationCount = 0
consecutiveFailures = 0

---

Main Loop

LOOP:
  1. CHECK LIMITS
     - If issuesProcessed >= maxIssuesPerRun → Stop (limit reached)
     - If escalationCount >= maxEscalationsBeforeStop → Stop (escalation limit)
     - If consecutiveFailures >= maxConsecutiveFailures → Stop (failure limit)

  2. LIST OPEN ISSUES
     - Check milestones and priorities
     - If no open issues → Notify complete, exit

  3. SELECT NEXT ISSUE
     - Follow latest implementation-plan file if exists
     - Otherwise use selection criteria (see below)
     - Check dependencies (skip if blocked)

  4. SPAWN /auto-work-issue SUBAGENT
     - Use Task tool to spawn: "Run /auto-work-issue {issue_number}"
     - Wait for completion
     - Parse output for result

  5. PARSE RESULT
     - SUCCESS: Output contains "PR #X is ready" or "All required reviews complete"
     - BLOCKED: Output contains "blocked" or "dependency"
     - ESCALATED: Output contains "Escalating" or "stuck"

  6. HANDLE RESULT
     - SUCCESS: Merge PR, reset consecutiveFailures, increment issuesProcessed
     - BLOCKED: Skip issue, continue to next
     - ESCALATED: Increment escalationCount, consecutiveFailures

  7. CONTINUE
     - Report progress
     - Loop back to step 1

END LOOP