Use when implementing any non-trivial feature, integration, refactor, or bugfix in this repo where there is a risk of skipping analysis/design/planning or skipping tests due to time pressure, scope pressure, or sunk-cost pressure.
Use when starting a new repo or onboarding an AI agent and you need to create or overhaul an AGENTS.md so work stays safe, testable, and consistent across environments (e.g., Godot + SDK + demo apps), especially when the project has strict tooling, sandboxes, or multiple "products" in one codebase.
适用于需要 PRD/Spec + 版本化执行计划（docs/plan/vN-*）+ 端到端验证的项目。通过严格的 PRD ↔ 计划 ↔ 测试 ↔ 代码追溯链，防止文档漂移与返工，直到愿景与实现之间无差异、无保留、无妥协。
Use when Codex repeatedly prompts for the same safe actions (running tests, reading logs, git operations) and the user wants those already-approved actions persisted into Codex approval config (trust levels + allow rules) so future sessions don’t require re-approval.
Render and view webpages using a headless browser (Playwright) to fetch JS-rendered HTML, extract visible text, and optionally save full-page screenshots. Use when a user asks to “无头浏览器打开/查看网页”, needs the rendered DOM instead of raw curl HTML, or wants a screenshot of a page.
End-to-end source code repository deep dive and review. Use when you need to understand a repo’s architecture, core abstractions, entrypoints and execution flow, extension points, and operational/developer setup; and when you must deliver a structured report (Markdown + standalone HTML) with diagrams (Mermaid), actionable recommendations, and a scored evaluation.
Use when a single file/class has grown into tightly coupled “spaghetti” (mixed UI, IO, state, input, gameplay), making changes risky, tests hard to write, and refactors frequently break behavior—especially under strict type/warning-as-error environments like Godot 4.x.