Use when a Formax repository task is non-trivial and you should analyze goals, non-goals, boundaries, data/type/interface impact, contract impact, test strategy, and whether an xhigh subagent is actually needed before writing a todo or code.
Use when working on Formax code changes and you need a disciplined dev loop: keep a single mainline task, avoid scope drift, run only targeted tests (no coverage), avoid partial staging (MM), run mandatory review before commit, include an incremental optimization check, and keep commits small and reviewable.
Use when implementing or debugging Formax permissions/policy/approval behavior and UI (allow/ask/deny rules, workspace boundaries, approval prompts, and /permissions overlay parity with Claude Code).
Use when preparing a Formax code handoff: selecting files, generating repomix bundles, and writing a high-quality prompt for WebGPT or another coding agent with clear constraints and validation scope.
Use when code has gone through repeated rework and may contain redundant logic, style drift, or tangled structure. Trigger when user asks for "返工收敛", "cleanup-pass", or requests a focused cleanup pass that reduces churn without changing behavior.
Use when the user wants a reusable rolling execution pattern (README + TODO-INDEX) like plans/web-reference-react-refactor, and wants to run delivery in small validated loops.
Use when adding or modifying generator scripts that write files; keeps generated artifacts out of docs/ and aligns script defaults with ownership paths plus CI gate updates.
Use when implementing or modifying behavior that must stay consistent across TUI and Web (mode/input/tool/replay/order). Require canonical semantics first, then TUI/Web adapters, then renderer-specific UI.
Use when we want to turn a just-finished Formax workflow (e.g. commands, overlays, tools, hooks, permissions, UI parity) into a reusable Codex Skill under .codex/skills, including scaffolding, guardrails, and the minimum test checklist.
Use when adding/modifying Formax slash commands (/permissions /agents /hooks /todos etc), especially output rendering (command sublines), overlay dismiss messages, and whether outputs are injected into the next LLM turn.
Use when designing, refactoring, or debugging Formax system prompt assembly, system-reminder injection, or skills/deferred-tool prompt exposure semantics; mandatory for any change involving FORMAXDEFERREDTOOLEXPOSURE and request-payload parity checks.
Implement or refactor Formax tool transcript UI using the Tool UI Blocks (C-lite) pattern (ToolUiBlocks renderer + blocks presenters) to avoid touching many tool presenter files; use when adjusting ⏺/⎿ spacing, indent rules, or migrating additional tools to blocks presenters with targeted Ink/Vitest tests and Codex…
Use for all frontend and web UI tasks by default to apply a business-agnostic desktop web visual style system with medium-strength guardrails; if an existing design system is present, follow it first and use this skill only as a style harmonizer.
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
Use after receiving WebGPT or another LLM review on a Formax todo and before sending the todo back for another pass. Produce a concise handoff response that says what we adopted, what we reject or question, and what the reviewer should specifically re-evaluate in docs/todolist.md.
Use after task analysis when a non-trivial Formax repository task needs a structured docs/todolist.md with [x]/[ ] items, definitions-before-UI ordering, explicit non-goals, tests, and loop-ready execution slices.