| name | subagent-execute |
| description | Executes a pre-planned subagent orchestration project by handing off tasks to specialized worker agents, monitoring reports, and adapting the plan. Use when asked to "execute a plan", "run subagents", "start execution", "hand off tasks", or when given a project name to execute against an existing docs/projects/<PROJECT-NAME>/ plan folder. |
Subagent Plan Execution
An orchestration skill for executing pre-planned projects by handing tasks to specialized worker agents, monitoring their completion reports, and adapting the plan as work progresses.
When to Use This Skill
- User invokes
/subagent-execute followed by a project name or phase reference
- A plan already exists in
docs/projects/<PROJECT-NAME>/ and is ready to execute
- User wants to hand off one or more task handoff documents to worker agents
- User wants to resume a partially-completed project by checking reports and continuing
Prerequisites
- A completed project plan in
docs/projects/<PROJECT-NAME>/ (created via /subagent-plan)
- At least one task handoff file in
docs/projects/<PROJECT-NAME>/tasks/
Complete Execution Process
Step 1: Read Required Conventions
Before executing, read these files:
.github/orchestration/SUBAGENT_FILE_CONVENTIONS.md ← File structure and naming rules
.github/orchestration/SUBAGENT_ORCHESTRATOR_GUIDE.md ← System overview
Step 2: Load the Project Plan
Read the master plan and current phase to understand state:
docs/projects/<PROJECT-NAME>/<PROJECT-NAME>-MASTER-PLAN.md
docs/projects/<PROJECT-NAME>/phases/<PROJECT-NAME>-PHASE-<##>-<NAME>.md
Identify:
- Current phase number
- Completed tasks (check
reports/ folder for existing *-REPORT.md files)
- Next task to execute
- Task dependencies
Step 3: Select Next Task
Determine the next task based on:
- Task sequence in the phase plan
- Prerequisites completed (check for task reports)
- No file conflicts with concurrent tasks
- Dependencies satisfied
If all Phase N tasks are complete → move to Phase N+1.
Step 4: Prepare Task Handoff
Read the task handoff document from tasks/:
Verify the task is ready:
Step 5: Hand Off to Worker Agent
Present the task handoff to the user for execution:
## 🎯 Task Ready for Execution
**Task**: <PROJECT-NAME>-TASK-<##>-<###>-<NAME>
**Assigned Agent**: Coding Wizard
**Priority**: [High/Medium/Low]
### Task Handoff Document
[Display or link to complete task handoff document]
### Instructions for Worker Agent
1. Read the complete task handoff document
2. Execute all task requirements
3. Write completion report following SUBAGENT_REPORT.md
4. Return report file path when complete
---
**To proceed**: Switch to **[Agent Name]** and provide the task handoff document.
Step 6: Monitor and Adapt (After Task Completion)
When the worker agent returns a report:
-
Read the Report at docs/projects/<PROJECT-NAME>/reports/<PROJECT-NAME>-TASK-<##>-<###>-REPORT.md
-
Verify Completion:
-
Handle Outcomes:
COMPLETE → Mark task done, identify next task, continue
PARTIAL → Review blockers, create follow-up task if needed, adjust plan
BLOCKED → Assess severity, create unblock task or adjust sequence
-
Handle Emerging Work — if worker discovers bugs or failures, create separate repair tasks rather than expanding original scope
-
Update Planning Docs — mark completed tasks, update master plan progress, adjust estimates
Step 7: Continue Execution Loop
Repeat Steps 3–6 until the phase or project is complete, or a blocking issue requires human intervention.
Execution Modes
Sequential (Default)
Execute tasks one at a time: hand off → wait for report → review → hand off next.
Best for tasks with dependencies or when learning-as-you-go.
Batch (Optional)
Hand off 2–3 non-conflicting tasks simultaneously to different agents.
Best for parallel components or independent modules with no shared files.
Quality Checks
Before each handoff:
After each report:
Progress Communication Format
## 📊 Execution Status
**Project**: <project-name>
**Current Phase**: Phase <##> - <name>
**Progress**: [X] of [Y] tasks complete
### Completed Tasks
✅ <PROJECT-NAME>-TASK-<##>-<###>-<NAME> (Agent: [Name])
### Current Task
🔄 <PROJECT-NAME>-TASK-<##>-<###>-<NAME>
Assigned to: [Agent Name]
Status: [In Progress | Ready for Handoff]
### Upcoming Tasks
⏳ <PROJECT-NAME>-TASK-XX-XXX-<NAME> (Waiting on: [prerequisites])
---
### Next Action
[Specific instruction for what to do next]
Emerging Issues
When workers report problems, create dedicated fix tasks:
| Issue | Task Name Pattern | Assign To |
|---|
| Bug found | <PROJECT-NAME>-TASK-XX-XXX-BUG-FIX | Coding Wizard |
| Tests failing | <PROJECT-NAME>-TASK-XX-XXX-TEST-FIX | Coding Wizard |
| Refactoring needed | <PROJECT-NAME>-TASK-XX-XXX-REFACTOR | Appropriate specialist |
References