RPI Quick

Goal

Coordinate one task through evidence, planning, execution, review, and explicit follow-up without duplicating the phase skills' detailed responsibilities.

Flow

1. Assess research readiness from caller-supplied research, task details, decisions, and plan inputs.
   1. Activate rpi-research only when evidence is missing, stale, contradictory, insufficient for planning, or when complexity, uncertainty, dependencies, risk, or a decision-critical question warrants investigation. Record Research disposition executed and consume the primary artifact's Planning Readiness.
   2. When evidence is adequate, record disposition reused or satisfied-and-skipped with the evidence that supports it.
   3. Apply the rpi-research continuation contract. Continue to Plan without another stage-start command only when either an executed Research primary artifact records Planning Readiness Ready, or reused or satisfied-and-skipped evidence is adequate, and all applicable gates pass, blockers clear, and required confirmations are explicit.
   4. When Research is Blocked, Needs clarification, or Not ready, or another transition requirement is not met, stop in Research and record the blocker or next action.
2. Run rpi-plan to create or revise the task-centered Markdown plan. Its rpi-plan-critique gate is internal to planning and returns its disposition to the planning parent.
3. Run rpi-implement for approved Pxx and Pxx-Txx work. Consume its return, including completed and remaining markers, validation coverage, blockers, plan updates, follow-up items, and readiness or the reason work is awaiting a significant or divergent user decision.
4. Run rpi-review once after Implementation returns and no affected work awaits a user decision. It uses one phase-and-task-matched review subagent, or an unnamed general-purpose subagent when no suitable specialist exists, to build the record from the current plan, critique, changes record, and validation evidence. Standard completely assesses the material acceptance boundary while minimizing elapsed work. The rpi-quick parent owns final outcome, route dispositions, and continuation. Use agent-owned Review decisions and skip item questions unless the user explicitly requested the Review Item Walkthrough. Record builder execution, final Review execution, outcome, and route decisions separately.
5. Follow-up: route defects, decision gaps, research gaps, and residual work to their correct next destination.

When Review finds open work, route it to the appropriate later stage or distinct follow-up item. Do not execute it or run Review again inside the current lifecycle.