Use at the start of MATERIAL or PROGRAM work when the user has a rough request, unclear success criteria, hidden trade-offs or missing decisions. Ask only high-impact questions, recommend defaults and produce a better execution brief.
Use when a project needs persistent storage and no database technology has been chosen yet, or an existing choice is being reconsidered. Selects a database with stated trade-offs, then designs the schema and API surface together against that choice. Do not…
Use for new UI, redesign, frontend implementation, screen or component review, responsive work, accessibility, generic AI appearance, or usability problems. Do not use for backend-only work.
Use when building dashboards, KPI tiles, management reports, or reconciliation views -- covers KPI breakdown completeness, the three-layer reporting model, and locale-correct currency/number formatting (symbol, digit grouping, unit scale) applied uniformly…
Use for design direction, brand, colour, typography, layout, visual style, animation, motion, responsiveness, design tokens, or creation/update of DESIGN.md. Also use when UI looks generic or inconsistent. Do not use for backend-only work.
Use during DISCOVER, immediately after inspecting the repository, to check whether the deterministic phase plan actually fits the task. Revise it with replan-phases when it does not. Do not use to invent phases outside the fixed ten-phase catalog.
Use when a domain problem may benefit from an existing AI agent implementation, including the 500+ AI Agents catalogue. Discover blueprints, inspect the exact linked repository and recommend reuse, adaptation or rejection. Never deploy a catalogue entry…
Use when a plan contains independent research, implementation or review tasks that benefit from fresh-context subagents or parallel execution. Do not parallelise tasks that modify the same state without a merge protocol.