# Ae Sdd Plan

> Research, plan, and prepare for implementation

- Skill: `shanepadgett/ae-sdd-plan` (Agent Skill)
- Install (CLI): `npx skillmds@latest add shanepadgett/ae-sdd-plan`
- Raw SKILL.md: https://api.skillmd.com/api/skills/shanepadgett/ae-sdd-plan/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: shanepadgett (https://skillmd.com/u/shanepadgett)
- Updated: 2026-09-10
- Page: https://skillmd.com/skills/shanepadgett/ae-sdd-plan

---


# Plan

Research the codebase and create an implementation plan. Full lane plans one task at a time; vibe/bug combines discovery + tasking + planning into one pass.

## Required Skills

- `spec-driven-development` (state/task management, lane detection)
- `research`

## Inputs

> [!IMPORTANT]
> Resolve the change set by running `ls changes/ | grep -v archive/`. If exactly one directory exists, use it. Only prompt the user when multiple change sets are present.

## Instructions

1. Load `spec-driven-development` skill and read state from `changes/<name>/state.toml`. Apply state entry check per skill guidelines.

2. Read tasks from `changes/<name>/tasks.toml` if full lane.

3. **Full lane**:
   - Confirm the active phase is `plan`
   - Determine task via CLI, not manual edits:
     - First check `ae sdd task current [name]`
     - If no current task, run `ae sdd task start [name]` to move next pending task to `in_progress`
   - Read current task details and all `thoughts/` (discovery outputs)
   - Research codebase for patterns, paths, integration, tests

4. **Vibe/Bug lane**:
   - Read `thoughts/` if they exist
   - Research codebase patterns, changes needed, risks
   - This is discovery + tasking + planning in one pass

5. **Dialogue**: Before writing, summarize research findings. If a path is clear, present your recommendation with reasoning. If you see questions, trade-offs, or risks, discuss them with opinions and recommendations. Collaborate through back-and-forth until direction is right.

6. **Plan contents** (all lanes):
   - Objective/Goal (one sentence, direct)
   - Requirements copied from task/spec as bullets
   - Research findings as bullets: paths, patterns, risks, constraints
   - Steps as an ordered list with exact file paths and concrete changes
   - Validation plan at the end in checkbox format

7. **Validation plan format**:
   - Place at the end of the plan
   - Two checkbox sections:
     - Agent validations (during/post implementation)
     - User validations (optional unless required)

8. **Full lane output**:
   - Create `changes/<name>/plans/<NN>.md`
   - Include all required sections
   - Steps must be exhaustive enough for direct execution

9. **Vibe/Bug lane output**:
   - Create `changes/<name>/plan.md`
   - Include all required sections
   - Include code or pseudocode for complex logic

10. Do not update phase status in this command. When the user wants to proceed, suggest `ae-sdd-next <name>` to advance into `implement`.

## Examples

**Full lane planning with dialogue:**

```text
Input: "password-reset" (full lane, task 2 pending)
Output: "Validated tasks—no in-progress tasks. Started task 2. Read task 2 and thoughts/. 
        Research shows extending auth/login/validator.ts fits DRY. 
        One concern: validator is getting large. Address now or focus on reset?"
       User: "Stay focused on reset. Show me the plan."
       Output: "Created plans/02.md with exhaustive breakdown: file-by-file changes, 
                code snippets for validator.js, auth/routes.js, tests, validation checklist."
```

**Vibe lane with clear recommendation:**

```text
Input: "bug-fix" (vibe lane)
Output: "Read thoughts/. Found consistent fix pattern in router/middleware.ts. 
        Applying that keeps codebase uniform. Created plan.md with exhaustive breakdown: 
        exact changes to middleware.ts, code snippet, test updates, validation steps."
```