# Plan Task (Architectural)

> Use when the user asks for a plan or the task is complex/ambiguous. Enforces AGENTS.md workflow and encourages loading other relevant skills.

- Skill: `majiayu000/plan-task-architectural` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds add majiayu000/plan-task-architectural`
- Raw SKILL.md: https://api.skillmd.com/api/skills/majiayu000/plan-task-architectural/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Productivity
- Author: majiayu000 (https://skillmd.com/u/majiayu000)
- Updated: 2026-09-09
- Page: https://skillmd.com/skills/majiayu000/plan-task-architectural

---


# Plan Task (Architectural)

Use this skill to produce a **fast, phased plan** consistent with `AGENTS.md`.

## When to use

- The user asks for a plan ("plan this", "how would you…", "phased plan").
- The task spans multiple files/subsystems.
- The requirements are ambiguous or risky.

## Protocol

1. Read: `architecture.md`, `README.md`, and the relevant `docs/architecture/*.md`.
2. Load supporting skills from `skills/` (see `skills/README.md`).

## Recommended output structure

### 1) Understanding & Mode

- 1–2 sentence summary of the request.
- Task mode (A/B/C/D/E) and what that implies (plan-only vs implement).

### 2) Investigation

- Files read (only what you opened).
- Current behavior/state (brief).
- Constraints (contracts, UX rules, limits).

### 3) Plan (phased)

Use 2–3 phases when helpful:

- Phase 1: preparation / refactor
- Phase 2: core change
- Phase 3: verification + cleanup + docs

For each phase: **what changes**, **files to touch**, **risks**.

### 4) Verification & Docs

- Commands to run (see `skills/verify_changes/SKILL.md`).
- Docs to update (see `skills/sync_docs/SKILL.md`).