How this skill is triggered — by the user, by Claude, or both
Slash command
/clear-thought-ultraplan:ultraplan-ctThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
---
references/output-format.mdreferences/plan-writing.mdreferences/prd-writing.mdreferences/research-protocol.mdreferences/review-checklist.mdreferences/thinking/collaborative-reasoning.mdreferences/thinking/debugging-approaches.mdreferences/thinking/decision-framework.mdreferences/thinking/design-patterns.mdreferences/thinking/mental-models.mdreferences/thinking/metacognitive-monitoring.mdreferences/thinking/programming-paradigms.mdreferences/thinking/scientific-method.mdreferences/thinking/sequential-thinking.mdreferences/thinking/structured-argumentation.mdreferences/thinking/visual-reasoning.mdreferences/understand-protocol.mdreferences/update-protocol.mdreferences/validate-protocol.mdreferences/xml-task-format.mdSingle-command planning pipeline: /ultraplan-ct [idea]
Orchestrates 6 phases with integrated thinking frameworks: UNDERSTAND --> RESEARCH --> PLAN --> REVIEW --> VALIDATE --> OUTPUT
Also supports: /ultraplan-ct update and /ultraplan-ct status
This skill enhances UltraPlan with 11 structured thinking frameworks from Clear Thought, applied at specific moments in each phase:
| Phase | Thinking Frameworks | Purpose |
|---|---|---|
| 1. UNDERSTAND | Mental Models + Metacognitive Monitoring | Decompose idea before questioning, assess knowledge gaps |
| 2. RESEARCH | Scientific Method + Decision Framework | Form testable hypotheses, evaluate options with weighted criteria |
| 3. PLAN | Design Patterns + Programming Paradigms + Visual Reasoning + Structured Argumentation | Select patterns, choose paradigms, create diagrams, justify decisions |
| 4. REVIEW | Collaborative Reasoning + Debugging Approaches + Metacognitive Monitoring | Multi-persona review, pre-mortem analysis, confidence calibration |
| 5. VALIDATE | Scientific Method + Sequential Thinking | Verify hypotheses, systematic tracing |
| 6. OUTPUT | Visual Reasoning + Mental Models | Architecture diagrams, Pareto prioritization |
Thinking framework references are in: {plugin_root}/skills/ultraplan-ct/references/thinking/
BEFORE doing anything else, follow these steps in exact order:
Parse the user's input to determine which command was invoked:
/ultraplan-ct [idea text] --> Main pipeline (go to Step 2)/ultraplan-ct update --> Update mode (jump to "UPDATE COMMAND" section at bottom)/ultraplan-ct status --> Status mode (jump to "STATUS COMMAND" section at bottom)Print this immediately:
================================================================
ULTRAPLAN CT: Clear Thought Enhanced Planning
================================================================
UNDERSTAND --> RESEARCH --> PLAN --> REVIEW --> VALIDATE --> OUTPUT
Your idea will go through 6 phases with structured thinking:
Phase 1: UNDERSTAND - First Principles analysis + discovery (40-70 questions)
Phase 2: RESEARCH - Hypothesis-driven parallel research
Phase 3: PLAN - PRD + technical plan with design patterns & diagrams
Phase 4: REVIEW - Multi-persona review + pre-mortem debugging
Phase 5: VALIDATE - Systematic requirement tracing & verification
Phase 6: OUTPUT - Visual summaries + Pareto-prioritized guide
11 thinking frameworks active. All output saved to: .ultraplan/
================================================================
Determine {plugin_root} by locating the directory containing this SKILL.md file. The plugin root is two levels up from the SKILL.md location (i.e., the clear-thought-ultraplan/ directory that contains skills/, agents/, hooks/, etc.).
Store {plugin_root} for use throughout the workflow. All reference file paths use this variable.
Check if .ultraplan/ directory exists in the current working directory.
If .ultraplan/ does NOT exist:
.ultraplan/ directory and .ultraplan/sections/ subdirectory.ultraplan/STATE.md using the template from {plugin_root}/skills/ultraplan-ct/templates/state.mdIf .ultraplan/ DOES exist:
.ultraplan/STATE.md to determine current position================================================================
EXISTING PLAN DETECTED
================================================================
Resuming from: Phase {N}/6 - {PHASE_NAME}
Last activity: {last_activity from STATE.md}
To start fresh, delete the .ultraplan/ directory and re-run.
================================================================
.ultraplan/DISCOVERY.md exists AND is complete (all 9 categories) --> Skip Phase 1.ultraplan/RESEARCH.md exists --> Skip Phase 2.ultraplan/sections/ contains section files --> Skip Phase 3.ultraplan/PLAN.md contains "Review Notes" section --> Skip Phase 4.ultraplan/VALIDATE.md exists --> Skip Phase 5.ultraplan/SUMMARY.md exists --> Already complete, show completion bannerIf DISCOVERY.md exists but is incomplete (not all 9 categories covered), resume Phase 1 from the next uncovered category.
If PRD.md exists but no section files exist, resume Phase 3 from Step 3b (technical plan generation).
If this is a fresh run (not resume), verify the user provided an idea:
/ultraplan-ct with no idea text:================================================================
ULTRAPLAN CT: Idea Required
================================================================
Please provide your idea in plain English. Examples:
/ultraplan-ct I want to build a recipe sharing app
/ultraplan-ct Build me a task management tool with team features
/ultraplan-ct Create an online store for handmade jewelry
Just describe what you want to build - no technical knowledge needed!
================================================================
Do not continue. Wait for user to re-invoke with an idea.
Purpose: Exhaustively gather requirements through structured questioning, guided by First Principles analysis.
Progress format: Phase 1/6: UNDERSTAND [===== ] 45% - Category: Edge Cases (6/9 categories)
{plugin_root}/skills/ultraplan-ct/references/understand-protocol.md for detailed question examples and category guidance.{plugin_root}/skills/ultraplan-ct/references/thinking/mental-models.md -- Apply First Principles Thinking to decompose the user's idea into fundamental truths before asking any questions.{plugin_root}/skills/ultraplan-ct/references/thinking/metacognitive-monitoring.md -- Assess what you know and don't know about this domain to prioritize deeper exploration.Before asking any questions, apply two thinking frameworks:
First Principles Analysis:
Metacognitive Assessment:
Document this analysis at the top of .ultraplan/DISCOVERY.md under a "## Pre-Discovery Analysis" section. Use this analysis to prioritize which question categories need deeper exploration.
Scan the current project directory for existing code:
package.json, requirements.txt, go.mod, Cargo.toml, *.csproj, pubspec.yaml, Gemfile, pom.xml, build.gradle, or any src/, app/, lib/ directories.git/ directory, README files, configuration filesIf codebase exists:
If no codebase (greenfield):
Create .ultraplan/DISCOVERY.md using the template from {plugin_root}/skills/ultraplan-ct/templates/discovery.md. Include the Pre-Discovery Analysis from Step 1.0.
Execute the question loop across all 9 categories. For each category:
Categories (must cover ALL 9):
| # | Category | Focus |
|---|---|---|
| 1 | Core Requirements | What must this do? Minimum viable version? What's out of scope? |
| 2 | Users & Context | Who uses this? Skill level? Environment? Devices? |
| 3 | Integration Points | What systems does this connect to? Data flows? APIs? |
| 4 | Edge Cases | What happens when things go wrong? Boundary conditions? |
| 5 | Quality Attributes | Performance needs? Security? Reliability? |
| 6 | Existing Patterns | (If codebase) How do similar things work here? Conventions? |
| 7 | Preferences & Tradeoffs | Strong opinions? Simplicity vs flexibility vs performance? |
| 8 | Monetization & Business Model | Free/paid? Subscriptions? Ads? Revenue model? |
| 9 | Visual & UX Vision | How should it look? Feel? Layout? Key screens? Interactions? |
For each category:
AskUserQuestion callAskUserQuestion with multiple-choice options:
.ultraplan/DISCOVERY.mdQuestion count enforcement:
Early stop handling:
After all 9 categories are covered:
.ultraplan/DISCOVERY.md.ultraplan/STATE.mdIf .ultraplan/DISCOVERY.md exists when resuming:
Purpose: Investigate solutions using hypothesis-driven research with weighted evaluation.
Progress format: Phase 2/6: RESEARCH [=== ] 30% - Launching subagents...
{plugin_root}/skills/ultraplan-ct/references/research-protocol.md for detailed research guidance.{plugin_root}/skills/ultraplan-ct/references/thinking/scientific-method.md -- Form testable hypotheses before spawning research agents.{plugin_root}/skills/ultraplan-ct/references/thinking/decision-framework.md -- Prepare weighted criteria for evaluating options.Before extracting topics, apply the Scientific Method:
.ultraplan/DISCOVERY.mdSpawn 3 subagents simultaneously in a SINGLE message (all 3 Task tool calls in one response):
Subagent 1: Codebase Researcher -- Analyze existing code or do tech stack comparison Subagent 2: Web Researcher -- Search web for best practices and approaches Subagent 3: Docs Researcher -- Fetch library docs via Context7 MCP
IMPORTANT: Launch all 3 in a single message so they run in parallel.
After all 3 subagents complete, merge findings into .ultraplan/RESEARCH.md.
When comparing technology options:
Compare research findings against user's discovery answers. Present conflicts via AskUserQuestion. Get user approval to proceed.
After research is complete:
Update STATE.md and proceed to Phase 3.
If .ultraplan/RESEARCH.md exists: skip entirely to Phase 3.
Purpose: Create PRD and technical plan with design pattern selection and architecture visualization.
Progress format: Phase 3/6: PLAN - PRD Section 4/10: What It Does
{plugin_root}/skills/ultraplan-ct/references/prd-writing.md for PRD generation rules.{plugin_root}/skills/ultraplan-ct/references/plan-writing.md for technical plan generation rules.{plugin_root}/skills/ultraplan-ct/references/thinking/design-patterns.md -- Identify applicable patterns.{plugin_root}/skills/ultraplan-ct/references/thinking/programming-paradigms.md -- Select paradigms per section.{plugin_root}/skills/ultraplan-ct/references/thinking/visual-reasoning.md -- Plan architecture diagrams.Generate the PRD one section at a time, getting user approval for each section before moving to the next. Follow the 10-section structure (What We're Building, The Problem, Who It's For, What It Does, How It Should Feel, What It Connects To, What It Does NOT Do, How We'll Know It Works, Business Model, Risks & Concerns).
For the Risks & Concerns section:
{plugin_root}/skills/ultraplan-ct/references/thinking/structured-argumentation.mdRead {plugin_root}/skills/ultraplan-ct/references/xml-task-format.md for XML task schema.
Step 3b.0: Design Pattern Selection (NEW - Clear Thought Enhancement)
Before splitting sections:
Step 3b.1: Derive Sections from PRD
Step 3b.2: Write Section Index -- .ultraplan/sections/index.md
Step 3b.3: Write Individual Section Files -- Each with overview, risk, dependencies, TDD stubs, XML tasks
Step 3b.4: Create Architecture Visualization (NEW - Clear Thought Enhancement)
Apply Visual Reasoning:
Step 3b.5: Write Master Plan -- .ultraplan/PLAN.md with architecture diagrams included
Step 3b.6: Justify Key Decisions (NEW - Clear Thought Enhancement)
Apply Structured Argumentation for the top 3 most important technical decisions:
Update STATE.md and proceed to Phase 4.
.ultraplan/PRD.md exists but .ultraplan/sections/ is empty: resume from Step 3b.ultraplan/sections/ contains section files: skip to Phase 4Purpose: Multi-perspective critical review with pre-mortem analysis and confidence calibration.
Progress format: Phase 4/6: REVIEW [==== ] 40% - Running checklist: Security (4/8)
{plugin_root}/skills/ultraplan-ct/references/review-checklist.md for the 8-category quality checklist.{plugin_root}/skills/ultraplan-ct/references/thinking/collaborative-reasoning.md -- Set up review personas.{plugin_root}/skills/ultraplan-ct/references/thinking/debugging-approaches.md -- Prepare pre-mortem analysis.{plugin_root}/skills/ultraplan-ct/references/thinking/metacognitive-monitoring.md -- Prepare confidence assessment.Create 3 review personas:
| Persona | Expertise | Focus | Known Bias |
|---|---|---|---|
| Security Expert | AppSec, threat modeling | Auth, data protection, input validation | Over-secures, may add unnecessary complexity |
| User Advocate | UX research, accessibility | User flows, onboarding, error messages | Under-estimates technical constraints |
| Devil's Advocate | Systems architecture | Failure modes, edge cases, scalability | Finds problems without always providing solutions |
Each persona reviews the plan independently. Their findings are combined with the standard checklist.
Run the 8-category quality checklist against ALL plan documents:
| # | Category | What to Check |
|---|---|---|
| 1 | Completeness | Does every discovery answer map to a plan section? |
| 2 | Consistency | Do sections contradict each other? Naming consistent? |
| 3 | Feasibility | Can each task actually be implemented as described? |
| 4 | Security | Are auth, validation, access control addressed? |
| 5 | Scalability | Will this architecture handle growth? |
| 6 | Edge Cases | Are error states, empty states, boundaries handled? |
| 7 | User Experience | Does the plan deliver the UX vision? Accessibility? |
| 8 | Cost & Complexity | Is the plan over-engineered? |
Auto-fix where possible, flag issues needing user decision.
For each yellow/red risk section:
Present results to user in plain English with the multi-persona findings.
Ask 5-10 NON-OBVIOUS questions the user probably didn't think about.
Apply Metacognitive Monitoring:
Integrate all findings (checklist + personas + pre-mortem + user answers) into affected sections. Append "Review Notes" to PLAN.md including confidence ratings.
Update STATE.md and proceed to Phase 5.
If .ultraplan/PLAN.md contains a "Review Notes" section: skip to Phase 5.
Purpose: Systematic requirement tracing with hypothesis verification.
Progress format: Phase 5/6: VALIDATE [====== ] 60% - Building traceability matrix...
{plugin_root}/skills/ultraplan-ct/references/validate-protocol.md for traceability rules.{plugin_root}/skills/ultraplan-ct/references/thinking/sequential-thinking.md -- For systematic tracing.{plugin_root}/skills/ultraplan-ct/references/thinking/scientific-method.md -- For hypothesis verification.Apply Sequential Thinking:
Extract requirements from DISCOVERY.md, extract tasks from section files, build the mapping.
For each unmapped requirement, present options to user via AskUserQuestion.
For each unmapped task, verify with user.
Revisit hypotheses from Phase 2:
Write .ultraplan/VALIDATE.md with full traceability matrix and hypothesis verification results.
Present summary and get approval to proceed.
Update STATE.md and proceed to Phase 6.
If .ultraplan/VALIDATE.md exists: skip to Phase 6.
Purpose: Produce final deliverables with visual summaries and prioritized execution guide.
Progress format: Phase 6/6: OUTPUT [======== ] 80% - Writing summary...
{plugin_root}/skills/ultraplan-ct/references/output-format.md for file format specifications.{plugin_root}/skills/ultraplan-ct/references/thinking/visual-reasoning.md -- For summary diagrams.{plugin_root}/skills/ultraplan-ct/references/thinking/mental-models.md -- For Pareto prioritization.Ensure every output file exists and is internally consistent.
Create .ultraplan/SUMMARY.md using the template, PLUS:
Visual Enhancement:
Pareto Enhancement:
Update with completion status and full session history.
Print the completion banner:
================================================================
ULTRAPLAN CT COMPLETE
================================================================
Your plan is ready! Enhanced with 11 thinking frameworks.
.ultraplan/
PRD.md - Your product requirements (plain English)
PLAN.md - Technical plan + architecture diagrams
RESEARCH.md - Research findings + decision matrices
DISCOVERY.md - Pre-discovery analysis + Q&A transcript
VALIDATE.md - Traceability matrix + hypothesis verification
STATE.md - Session state (for resume/updates)
SUMMARY.md - One-page cheat sheet with diagrams
sections/
index.md - Section manifest
section-01-* - {section 1 name} [{risk color}]
section-02-* - {section 2 name} [{risk color}]
...
Total: {N} sections, {M} tasks
Thinking frameworks applied: 11
----------------------------------------------------------------
WHAT TO DO NEXT
----------------------------------------------------------------
CRITICAL PATH (highest-value sections):
{list of Pareto-identified sections}
To start building:
Give the .ultraplan/ folder to any AI coding tool and say:
"Read .ultraplan/sections/index.md and execute section 1"
To update this plan:
Run: /ultraplan-ct update
To view the summary:
Read: .ultraplan/SUMMARY.md
================================================================
If .ultraplan/SUMMARY.md exists: the plan is already complete. Show the completion banner again.
Triggered by: /ultraplan-ct update
.ultraplan/ directory exists. If not, tell user to run /ultraplan-ct first.{plugin_root}/skills/ultraplan-ct/references/update-protocol.md for detailed update behavior.AskUserQuestionTriggered by: /ultraplan-ct status
Read .ultraplan/STATE.md and present current status with progress bar, file manifest, and session history.
CRITICAL: When resuming after context compaction:
{plugin_root} by locating this SKILL.md fileAskUserQuestion with multiple-choice optionsPlain English everywhere. The target user is a no-coder. Never use technical jargon in user-facing text.
AskUserQuestion for ALL user input. Always provide multiple-choice with a recommended first option.
Auto-save after every batch/step. Never lose progress.
Progress indicators after every meaningful step.
Non-destructive updates. /ultraplan-ct update NEVER regenerates the full plan.
Sections are the atomic unit. Each section is self-contained with XML tasks, TDD stubs, risk rating, and dependency info.
Discovery is exhaustive. 40-70 questions across 9 categories.
Research is parallel. Always launch 3 subagents simultaneously.
PRD approval is incremental. One section at a time.
Review is multi-perspective. 3 review personas + 8-category checklist + pre-mortem debugging.
Validation is bidirectional. Requirements-to-tasks AND tasks-to-requirements.
State is always saved. STATE.md is the source of truth for resume.
Thinking frameworks are applied, not just read. When a phase says to use a framework, produce the structured output format defined in the framework reference file.
npx claudepluginhub ankurjain1121/dev-workflow-skills --plugin clear-thought-ultraplanCreates structured plans for multi-step tasks including software features, implementations, research, or projects. Deepens plans via interactive sub-agent reviews.
Strategic planning with auto-calibrated detail, decision rationale, and dependency ordering. Use when starting a new feature, bug fix, refactor, or any non-trivial work. Produces a plan document with tasks, reasoning, and acceptance criteria. Triggers: plan, planning, create plan, implementation plan, feature plan, work plan.
Generates atomic PLAN.md files for hierarchical project planning in solo agentic dev with Claude. Covers briefs, roadmaps, phases; includes tasks, verification, checkpoints, success criteria.