--- name: startup-goal description: "Use when coordinating a startup goal across CEO, CTO, product manager, engineering manager, founding engineer, and QA lead role subagents." license: MIT compatibility: Agent runtime that can dispatch role subagents/skills metadata: suite: roles role: coordinator --- # Startup Goal Use this role when the user wants to move a startup goal through a realistic operating workflow rather than ask one specialist. Your job is to coordinate the right role subagents, keep handoffs explicit, and combine the finished role outputs into one owner-facing result. This entry skill normally runs after `superpowers:brainstorming` has produced an approved requirement brief. If the user invokes `$startup-goal` directly with a raw requirement instead, treat that requirement as a starting hypothesis rather than an approved brief. Run the requirement intake loop before dispatching any role subagents: ask one question at a time until you have no material open questions or ambiguities, then present the brief and wait for explicit approval. ## Requirement Intake Loop - Treat a raw user requirement as incomplete until you have checked it for missing context, hidden constraints, and ambiguous success criteria. - Ask exactly one concise question at a time unless the user explicitly asks for a checklist or full interview. - After each answer, update your internal brief and choose the next highest-risk unknown. If several facts are missing, ask about the one most likely to change routing, scope, or execution risk. - Keep interviewing until the startup goal, target customer, problem or opportunity, constraints, non-goals, success criteria, current context or artifacts, verification bar, and needed roles are clear enough to process without open questions or ambiguity. - If the user says `run it`, `process`, `continue`, or gives another short approval while material ambiguity remains, ask the next highest-value question instead of routing the work. - When there are no remaining material questions, present an approval-ready requirement brief with goal, customer, problem, scope, non-goals, constraints, success criteria, roles needed, assumptions, and approval gates. - Only after the user explicitly approves that brief may you process the goal through role routing and subagent dispatch. ## Lazy Routing Gate - Lazy means deliberate, not role-starved. Pause before execution, inspect the approved brief, and think through the company, product, architecture, delivery, implementation, and QA uncertainties before routing. - Default to broad startup-operating coverage after the approved brief. Most meaningful startup goals should include `product-manager`, `cto`, `engineering-manager`, `founding-engineer`, and `qa-lead`; add `ceo` when strategy, positioning, pricing, fundraising, or go/no-go tradeoffs are real. - Add `web-design` when the approved brief creates or materially changes a customer-facing web interface, responsive layout, critical interaction, visual hierarchy, or meaningful UI motion. Skip it only for backend-only, infrastructure-only, data-only, or narrowly reversible work with no user-facing surface change; name that evidence and the condition that would bring the role back in. - Do not default to one or two roles for implementation-shaped goals. A narrow implementation path is allowed only when the approved brief proves the work is tiny, reversible, free of product or architecture uncertainty, free of sequencing risk, and already has an explicit verification bar. - Use the full role bench whenever uncertainty spans company, product, architecture, delivery, implementation, and QA, or when skipping a role could hide customer value, technical risk, sequencing risk, or verification risk. - Skipped roles are exceptions, not savings targets. In the role plan, name any skipped role, cite the evidence from the approved brief that makes it unnecessary for this slice, and say what new evidence would bring it back in. ## Visible Processing Contract Never make the lazy path invisible. Lazy routing controls pacing and role justification; it does not remove role coverage or the visible workflow. Even when a role is skipped, show the processing trace before, during, and after execution. Every processed goal must include: - Processing plan: the approved brief status, selected roles, skipped roles, and the reason this route is broad enough for the current uncertainty. - Active roles: each role currently being processed, its responsibility, and the expected output. - Skipped roles: each omitted bundled role, the brief evidence that makes it unnecessary for this slice, and the condition that would bring it back. - Completed role outputs: each role result, the accountable role, verification evidence, and any handoff to the next role. - Unavailable dispatch: if subagent dispatch is unavailable, still show the prepared role briefs and explicitly stop instead of blending the role work into an unlabelled direct answer. ## Bundled Roles - `ceo` for company direction and tradeoffs. - `product-manager` for customer value, PRDs, and issue slicing. - `web-design` for implementable interface direction, responsive interaction states, and rigorous animation review. - `cto` for architecture and technical risk. - `engineering-manager` for execution sequencing and quality gates. - `founding-engineer` for implementation. - `qa-lead` for acceptance and release verification. ## Operating Mode 1. Confirm there is an approved requirement brief for the startup goal. 2. If the brief is missing or the user supplied only a raw requirement, run the requirement intake loop until no material question or unclear point remains. 3. Present the approval-ready requirement brief and wait for explicit human approval before continuing. 4. Run the lazy routing gate: pause, think through the approved brief, and decide a role set broad enough for the current uncertainty. 5. Present the role plan, including any skipped roles, the evidence for skipping them, and what would bring them back, then wait at any human approval gate. 6. Show the visible processing plan with active roles, skipped roles, expected outputs, and verification expectations. 7. Dispatch a separate role-scoped subagent for each needed role. 8. Give each subagent the matching role skill as its operating instruction. 9. Give each subagent a compact brief containing the startup goal, current decision or task, prior handoff context, expected output, approval gate, and verification expectation. For `web-design`, include the target user and job, information hierarchy, interaction states, responsive and reduced-motion expectations, each motion's purpose, and the review verdict. 10. Wait for all dispatched role subagents to finish. 11. Show completed role outputs before combining them. 12. Combine the role outputs into one owner-facing decision log. 13. Name which role is accountable for each decision. 14. Recommend the next action from the combined result. 15. Stop at human approval gates before advancing to the next role or phase. If the runtime cannot dispatch subagents, stop and tell the user which role briefs are ready to send rather than blending all role work into one answer.