-
Notifications
You must be signed in to change notification settings - Fork 2
Open
Labels
agent-drivenWork primarily executed by automated agentsWork primarily executed by automated agentsautomatedenhancementNew feature or requestNew feature or requestmixed-collaborationCollaborative work between agents and humansCollaborative work between agents and humansorchestration
Description
Configuration Parameters (Optional)
Customize iteration behavior by setting these parameters:
- max_tasks: 8
- scope_focus: Enabler tasks (blocking other work)
- priority_threshold: high
- mode: Execution mode
Objective
Execute an orchestration cycle to process pending tasks in the work queue following the file-based orchestration approach. Focus on Tasks that enable other tasks to progress.
Context
- Orchestration Framework: Implemented and production-ready (PR Add agent orchestrator and validation tooling #16)
- Approach: File-based orchestration (
.github/agents/approaches/file-based-orchestration.md) - Work Directory:
work/(inbox, assigned, done, logs, collaboration) - Agent Profiles:
.github/agents/*.agent.md - Directives:
.github/agents/directives/ - Mode:
/analysis-modefor task execution
Current Status
To be filled/check by Manager Mike before running iteration running work/scripts/template-status-checker.sh or manually:
bash work/scripts/template-status-checker.shManual alternative:
echo "Inbox: $(ls work/inbox/*.yaml 2>/dev/null | wc -l) tasks"
echo "Assigned: $(find work/assigned -name '*.yaml' 2>/dev/null | wc -l) tasks"
echo "Done: $(find work/done -name '*.yaml' 2>/dev/null | wc -l) tasks"Execution Instructions
1. Initialize
- Initialize as Manager Mike per
AGENTS.mdguidelines - Read
work/collaboration/AGENT_STATUS.mdfor current context - Verify orchestration framework operational status
2. Task Selection
- Check
work/inbox/for pending tasks - Run
python work/scripts/agent_orchestrator.pyto assign tasks - Identify top priority tasks:
- Priority order: critical > high > medium > normal
- Consider dependencies (tasks without dependencies, and enabling others first)
- Check POC/chain continuity (advance multi-agent chains)
- Take special care to avoid dependency loops! ( A -> B -> A)
3. Task Execution
- Delegate each task to appropriate custom agent
- Follow file-based orchestration lifecycle:
- Update status: assigned → in_progress
- Execute task per agent specialization
- Create artifacts as specified
- Update task YAML with result block
- Update status: in_progress → done
- Move task from
work/assigned/<agent>/towork/done/<agent>/- Important: Tasks must be moved to the agent-specific subdirectory, not directly to
work/done/root
- Important: Tasks must be moved to the agent-specific subdirectory, not directly to
- Create work log in
work/logs/<agent>/per Directive 014
4. Documentation & Reporting
- Create work log for each completed task (Directive 014)
- Update
work/collaboration/AGENT_STATUS.md - Create iteration summary (Manager Mike)
- Capture metrics: duration, tokens, artifacts, validation results
- Commit changes with
report_progressafter each task - Update PR description with iteration summary
- Post Manager Mike recap comment to PR
Success Criteria
Note: These checkboxes can be auto-populated by orchestrator post-execution using work/scripts/template-status-checker.sh --validate
- At least 1 task completed (or all remaining if fewer than 3)
- All work logs created per Directive 014
-
AGENT_STATUS.mdupdated with current iteration state - Iteration summary created by Manager Mike
- Metrics captured and reported
- Framework health assessed (completion rate, validation success, etc.)
- Zero validation errors or failures
- All artifacts committed and pushed
Deliverables
- Completed Tasks: Moved to
work/done/<agent>/subdirectories - Work Logs: Created in
work/logs/<agent>/ - Iteration Summary: Added to
work/collaboration/ - Updated Status:
AGENT_STATUS.mdreflects current state - Manager Recap: Comment posted to PR
- Metrics Report: Duration, tokens, artifacts, success rate
References
- File-Based Orchestration:
.github/agents/approaches/file-based-orchestration.md - AGENTS.md: Root orchestration protocol
- Directive 014: Work Log Creation (
.github/agents/directives/014_worklog_creation.md) - ADR-009: Orchestration Metrics Standard
- Agent Profiles:
.github/agents/*.agent.md - Previous Iterations: Check
work/collaboration/ITERATION_*_SUMMARY.md
Example Prompt
Execute an orchestration cycle following file-based orchestration approach:
1. Initialize as Manager Mike per AGENTS.md
2. Read work/collaboration/AGENT_STATUS.md
3. Run agent_orchestrator.py to assign pending tasks
4. Execute top X priority tasks by delegating to custom agents ( see max tasks configuration at the top )
5. Create work logs per Directive 014
6. Update AGENT_STATUS.md and create iteration summary
7. Post Manager Mike recap comment
Ensure:
- Tasks selected by priority (critical > high > medium)
- All work logs Directive 014 compliant
- Metrics captured per ADR-009
- Incremental commits with report_progress
- Framework health assessed
Metadata
Metadata
Assignees
Labels
agent-drivenWork primarily executed by automated agentsWork primarily executed by automated agentsautomatedenhancementNew feature or requestNew feature or requestmixed-collaborationCollaborative work between agents and humansCollaborative work between agents and humansorchestration