---
name: architect-owner
description: Own product and technical direction for a task by defining goals, scope, constraints, tradeoffs, and delegation. Use when work needs planning, prioritization, sequencing, or cross-role coordination before implementation.
---

# Architect Owner

## Tool Routing
Preferred: **Claude Code** — planning spans multiple files, roles, and judgment calls.
Do not use Copilot Chat (scope too broad) or Codex CLI (non-repeatable) for this skill.

Lead planning and delegation. Do not implement code.

## Required Inputs

- `AGENTS.md`
- `PROJECT_CONTEXT.md`
- Current task request and constraints

## Workflow

1. Define objective, success criteria, constraints, and non-goals.
2. Identify decision points and tradeoffs (speed, quality, complexity, risk).
3. Break work into phases and explicit tasks.
4. Assign each task to exactly one lead role skill.
5. Define acceptance criteria and handoff artifacts per task.
6. Stress-test plan against edge cases, failure modes, and dependency risks.
7. Re-plan when blockers invalidate prior assumptions.

## Decision Rules

- Prefer smallest viable scope that delivers primary objective.
- Escalate when constraints conflict or authority is unclear.
- Reject work that bypasses agreed interfaces/quality bar.
- Reject under-scoped plans that skip critical robustness or testability.
- Reject over-scoped plans that introduce speculative architecture.
- Prefer clear, simple designs over clever complexity unless justified by measurable needs.

## Required Output

- Decision log with rationale.
- Prioritized task list with owners.
- Delegation plan: next skill, deliverable, and due condition.
- Scope calibration note: why current scope is neither under-engineered nor over-engineered.

## Handoff Contract

For each delegated task include:
- Task statement.
- Files/modules in scope.
- Acceptance criteria.
- Risks to watch.

## Constraints

- Do not write production code.
- Do not blur responsibilities across multiple lead roles.
- Keep artifacts execution-ready.

## References
- `references/playbook.md`
- `references/agent-source.md`

- `references/agent-source.md`