dev-blueprint

version	1.0.0
name	dev-blueprint
description	Turn a one-line objective into a step-by-step construction plan for multi-session, multi-agent projects. Each step has a self-contained context brief so a fresh agent can execute it cold. Includes adversarial review, dependency graph, and parallel step detection. Use when the user requests a plan or roadmap for a complex multi-PR task, or describes work that needs multiple sessions. Do NOT use for tasks completable in a single PR or fewer than 3 tool calls.
disable-model-invocation	true
user-invocable	true
argument-hint	[project] "objective description"

Blueprint — Construction Plan Generator

Turn a one-line objective into a step-by-step construction plan that any coding agent can execute cold.

When to Use

• Breaking a large feature into multiple PRs with clear dependency order
• Planning a refactor or migration that spans multiple sessions
• Coordinating parallel workstreams across sub-agents
• Any task where context loss between sessions would cause rework

Do not use for tasks completable in a single PR, fewer than 3 tool calls, or when the user says "just do it."

How It Works

Blueprint runs a 5-phase pipeline:

1. Research — Pre-flight checks (git, gh auth, remote, default branch), then reads project structure, existing plans, and memory files to gather context.
2. Design — Breaks the objective into one-PR-sized steps (3-12 typical). Assigns dependency edges, parallel/serial ordering, model tier, and rollback strategy per step.
3. Draft — Writes a self-contained Markdown plan file. Every step includes a context brief, task list, verification commands, and exit criteria — so a fresh agent can execute any step without reading prior steps.
4. Review — Delegates adversarial review to a strongest-model sub-agent against a checklist and anti-pattern catalog. Fixes all critical findings before finalizing.
5. Register — Saves the plan, updates memory, and presents the step count and parallelism summary to the user.

Output Format

Plans are saved to plans/ (or docs repo if workflow.json exists):

# Blueprint: {objective}

**Created:** {date}
**Steps:** {N} ({M} parallel groups)
**Estimated PRs:** {N}

## Dependency Graph

```mermaid
graph TD
    S1[Step 1: Interface] --> S2[Step 2: Backend]
    S1 --> S3[Step 3: Frontend]
    S2 --> S4[Step 4: Integration tests]
    S3 --> S4
    S4 --> S5[Step 5: Cleanup]

Step 1: {title}

Context Brief

{Everything a fresh agent needs to know to execute this step. No reference to prior steps. Self-contained.}

Tasks

• {task 1}
• {task 2}

Verification

{commands to verify this step is done correctly}

Exit Criteria

• {criterion 1}
• {criterion 2}

Rollback

{How to undo this step if needed}

---

Step 2: {title}

...

## Key Features

- **Cold-start execution** — every step includes a self-contained context brief
- **Adversarial review gate** — plan reviewed by a strongest-model sub-agent
- **Branch/PR workflow** — built into every step, degrades to direct mode without git/gh
- **Parallel step detection** — dependency graph identifies parallelizable work
- **Plan mutation** — steps can be split, inserted, skipped, reordered, or abandoned

## Examples

/dev-blueprint my-api "migrate from Dapper to EF Core" /dev-blueprint my-app "add real-time notifications with SignalR" /dev-blueprint my-frontend "extract shared UI components into a library"

## Anti-Pattern Catalog

Plans are reviewed against these anti-patterns:

| Anti-pattern | Detection |
|-------------|-----------|
| **Big bang step** | Any step touching >10 files |
| **Hidden dependency** | Step B reads file only modified in step A, but no edge exists |
| **Missing verification** | Step has no verification commands |
| **Context leak** | Step references "as we did in step N" instead of being self-contained |
| **No rollback** | Destructive step has no undo strategy |
| **Over-parallelism** | Parallel steps share files |