ワンクリックで
blueprint
Transform feature descriptions into well-structured project plans
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Transform feature descriptions into well-structured project plans
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Create distinctive, production-grade frontend interfaces with high design quality. Use when building web components, pages, or applications. Includes framework-specific guidance for Tailwind, React, Vue, and Rails/Hotwire ecosystems.
Skill file structure, naming conventions, directory layout, frontmatter requirements, and invocation control. Use when creating skill files or slash commands to ensure correct format and validation.
Forces adversarial reasoning before committing to decisions. Triggers on architectural choices, approach selection, and planning phases to prevent premature commitment bias.
Verify LLM reasoning using Reverse Chain-of-Thought (RCoT) to detect overlooked conditions.
Autonomous task implementation - research, plan, build, review, fix, ship
Generate a complete favicon set (ICO, PNG variants, apple-touch-icon, web manifest) from a source image and integrate into the project's HTML layout. Use when user asks to generate favicons, set up PWA icons, or add an apple-touch-icon.
| name | blueprint |
| description | Transform feature descriptions into well-structured project plans |
| argument-hint | [feature description, bug report, or improvement idea] |
CRITICAL: This is a WORKFLOW SKILL, not Claude's built-in plan mode. Do NOT use EnterPlanMode or ExitPlanMode tools.
TASK_TRACKING = config_read("task_tracking.enabled", "false")
If TASK_TRACKING:
BLUEPRINT_WORKFLOW_ID = "blueprint-{timestamp}"
PHASE_TASKS = {}
PHASES = [
{num: 1, name: "Discovery", active: "Running discovery"},
{num: 2, name: "Research", active: "Researching codebase"},
{num: 3, name: "Research Validation", active: "Validating research"},
{num: 4, name: "Architecture", active: "Designing architecture"},
{num: 5, name: "Capture Learnings", active: "Capturing learnings"},
{num: 6, name: "Write Plan", active: "Writing plan"},
{num: 7, name: "Review Plan", active: "Reviewing plan"},
{num: 8, name: "User Decision", active: "Awaiting user decision"},
{num: 9, name: "Execution", active: "Executing plan"}
]
For each P in PHASES:
PHASE_TASKS[P.num] = TaskCreate(
subject: "Phase {P.num}: {P.name}",
activeForm: P.active,
metadata: {workflow: BLUEPRINT_WORKFLOW_ID, phase: P.num}
)
# Dependencies: phase 4 blocked by 1-3, then sequential 5→6→7→8→9
TaskUpdate(PHASE_TASKS[4], addBlockedBy: [PHASE_TASKS[1], PHASE_TASKS[2], PHASE_TASKS[3]])
TaskUpdate(PHASE_TASKS[5], addBlockedBy: [PHASE_TASKS[4]])
TaskUpdate(PHASE_TASKS[6], addBlockedBy: [PHASE_TASKS[5]])
TaskUpdate(PHASE_TASKS[7], addBlockedBy: [PHASE_TASKS[6]])
TaskUpdate(PHASE_TASKS[8], addBlockedBy: [PHASE_TASKS[7]])
TaskUpdate(PHASE_TASKS[9], addBlockedBy: [PHASE_TASKS[8]])
<feature_description> $ARGUMENTS </feature_description>
If empty: Ask user to describe the feature, bug fix, or improvement.
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[1], status: "in_progress")
Skill("blueprint-discovery")
Handles: Idea Refinement, Interview decision, Acceptance Criteria, Feature Classification.
Wait for output: discovery_result
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[1], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[2], status: "in_progress")
Skill("blueprint-research") feature_description, discovery_result
Handles: Toolbox resolution, lessons discovery, parallel research agents, spec review.
Wait for output: research_result
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[2], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[3], status: "in_progress")
Quick checkpoint before architecture. Catches misalignment early.
research_summary = summarize research_result:
- Key patterns found (2-3 bullets)
- Similar implementations discovered
- Potential constraints/blockers
AskUserQuestion:
question: "Research found: {research_summary}. Continue to planning?"
header: "Validate"
options:
- label: "Looks good"
description: "Proceed to architecture phase"
- label: "Missing something"
description: "Tell me what to research deeper"
- label: "Wrong direction"
description: "Let's revisit the requirements"
If "Missing something":
AskUserQuestion: "What should I dig into?"
→ Run targeted research agent
→ Append to research_result
→ Return to Phase 3
If "Wrong direction":
→ Return to Phase 1 (Discovery)
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[3], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[4], status: "in_progress")
Task(majestic-engineer:plan:architect):
prompt: |
Feature: {feature_description}
Research: {research_result.research_findings}
Spec: {research_result.spec_findings}
Skills: {research_result.skill_content}
Lessons: {research_result.lessons_context}
Wait for output: architect_output
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[4], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[5], status: "in_progress")
PRIMARY_DIR = extract main folder from architect_output.file_recommendations
AGENTS_MD = walk up from PRIMARY_DIR until AGENTS.md found (default: root)
If new patterns or decisions discovered:
Edit(AGENTS_MD, append patterns/decisions)
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[5], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[6], status: "in_progress")
Apply plan-builder skill.
Select template based on complexity:
Output: Write to docs/plans/[YYYYMMDDHHMMSS]_<title>.md
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[6], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[7], status: "in_progress")
Task(majestic-engineer:plan:plan-review):
prompt: "Review plan at {plan_path}"
Incorporate feedback and update plan file.
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[7], status: "completed")
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[8], status: "in_progress")
Check auto_preview: config_read("auto_preview", "false")
If true: Bash(command: "open {plan_path}")
AskUserQuestion:
question: "Blueprint ready at `{plan_path}`. What next?"
options:
- "Build as single task" → Phase 9a
- "Break into small tasks" → Phase 9b
- "Create as single epic" → Phase 9c
- "Deep dive into specifics" → Phase 8.1
- "Preview plan" → Read and display, return to Phase 8
- "Revise" → Ask what to change, return to Phase 6
- "Review and refine" → Apply document-refinement skill, return to Phase 8
IMPORTANT: After user selects, EXECUTE that action.
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[8], status: "completed")
If "Review and refine":
Apply document-refinement skill to {plan_path}
→ Auto-fix minor issues in plan file
→ Present refinement report to user
→ Return to Phase 8
AskUserQuestion: "What aspect needs deeper research?"
Run focused research:
Task(majestic-engineer:research:best-practices-researcher):
prompt: "Deep dive: {aspect} for feature {feature}"
Task(majestic-engineer:research:web-research):
prompt: "{aspect} - patterns, examples, gotchas"
Update plan with findings, return to Phase 8.
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[9], status: "in_progress")
Dispatch on user_choice from Phase 8:
9a. Single Task (user_choice == "Build as single task"):
Skill("build-task") "{plan_path}"
End workflow.
9b. Break Into Tasks (user_choice == "Break into small tasks"):
Task(majestic-engineer:plan:task-breakdown):
prompt: "Plan: {plan_path}"
AskUserQuestion:
question: "Tasks added to plan. Create in backlog?"
options:
- "Yes, create tasks" → For each task in plan: Skill("backlog-manager")
- "No, just the plan" → Skip
Then offer build:
AskUserQuestion:
question: "Start building?"
options:
- "Build all tasks now" → Skill("run-blueprint") "{plan_path}"
- "Done for now" → End workflow
9c. Single Epic (user_choice == "Create as single epic"):
Skill("backlog-manager")
Update plan with task reference, then offer build:
AskUserQuestion:
question: "Start building?"
options:
- "Build all tasks now" → Skill("run-blueprint") "{plan_path}"
- "Done for now" → End workflow
If TASK_TRACKING: TaskUpdate(PHASE_TASKS[9], status: "completed")
| Scenario | Action |
|---|---|
| Discovery skill fails | Ask user for AC manually, continue |
| Research skill fails | Continue with architect (degraded) |
| User skips validation | Continue to Phase 4 |
| Architect fails | Log error, ask user to provide approach |
| Plan-review fails | Continue with original plan |
| Task creation fails | Report error, ask user to create manually |
| Config read fails | Use default value |
| User cancels | End workflow gracefully |