Autonomous agent loop for executing, validating, and completing tasks. Handles state transitions, subtask management, and review cycles.
Resources
2Install
npx skillscat add dimitrigilbert/task-o-matic/task-orchestrator Install via the SkillsCat registry.
SKILL.md
Task Orchestrator
Goal: Drive tasks from todo to completed with strict quality gates.
Core Rules
- Validation First: NEVER complete a task without passing build/tests.
- Subtask Priority: BLOCKED if subtasks are incomplete.
- Status Truth: ALWAYS update status to reflect reality (
in-progresswhen starting). - Context Aware: ALWAYS read linked PRD requirements before execution.
The Loop
1. Claim & Context
Find work and lock it.
# Find next task
npx task-o-matic tasks get-next --status todo
# Set status (replace <ID> with the ID found above)
npx task-o-matic tasks status --id <ID> --status in-progress
# Load Context (Check for prdFile/prdRequirement)
npx task-o-matic tasks show --id <ID>2. Execute
Perform the work. Use available tools (opencode, edit, bash).
If task is complex, split it: npx task-o-matic tasks split --id $TASK_ID
3. Validate
Run project-specific checks.
# Example validation cycle
bun run check-types && bun run build && bun test4. Review & Fix
Triggers code-reviewer skill if validation passes.
If Review Fails: Create fix subtasks.
npx task-o-matic tasks create --parent-id $TASK_ID --title "Fix: <issue>" --effort smallLoop back to Step 1 for the fix subtask.
If Review Passes: Proceed to completion.
5. Completion
Only when:
- Code implemented
- Validation passed
- Review approved
- All subtasks completed
npx task-o-matic tasks status --id $TASK_ID --status completedHelper Scripts
scripts/check-status.sh <id>: View task status and subtask tree.