Automate a Manifest-bound delivery across Positioning, PRD, Spec, Impact, Plan, Test Plan, implementation, verification, and Candidate review by coordinating independent author/reviewer loops inside a human-approved Authority Envelope. Use when a user asks to…
Write implementation plans for features that have passed a human or independent artifact-review Spec Gate. Defines Phase breakdown, task sizing (XS/S/M), dependency graph, and estimated vs actual time tracking. Trigger: after a valid Spec Gate, before Test…
Use when writing a PRD, starting a new feature that needs requirements, or when user says 写PRD, 产品需求, feature requirements, 需求文档. Defines the 13-section PRD structure with mandatory observability requirements.
Use when PRD has passed a valid human or independent artifact-review Gate and technical spec is needed, or when user says 写Spec, 技术方案, tech spec, technical design, 架构设计. Defines the 12-section technical specification with mandatory architecture diagrams and…
Write test plans that define Critical User Journeys (CUJ), test pyramid structure, and mock strategy. Trigger: after a valid human or independent artifact-review Plan Gate, before Implementation.
Use as the ENTRY POINT when starting any new project, new feature, or complex task that needs a structured approach. Enforces the sequential gate flow Stage 00 (Positioning) → Stage 01 (PRD) → Stage 02 (Spec) → Stage 03 (Plan) → Stage 04 (Test Plan) before…
Apply the platform-neutral Tier, approval, evidence, and irreversible-action boundaries defined by Agent Engineering Workflow.
Use when multiple AI agents (Claude Code, Codex, OpenCode, etc.) are coding the same git repo in parallel, or when a single agent's run is long enough that other agents may commit during it. Enforces the 3-layer defense (Declaration / Isolation / Detection)…