Sequential Thinking

Break down complex problems into numbered, reflective thought sequences that can revise, branch, and verify hypotheses dynamically.

Core Principle

Complex problems require visible reasoning chains, not jumbled analysis.

When you work through multi-step problems, your brain benefits from explicit thought progression: “This led to that, which revealed this flaw, so I’ll revise my approach.” Sequential Thinking makes this process systematic with numbered thoughts, revision markers, and hypothesis verification.

When to Use

Always use for:

  • Multi-step problems with unclear scope
  • Debugging that requires hypothesis testing
  • Architecture decisions comparing alternatives
  • Analysis requiring course correction mid-thought

Especially when:

  • You discover new complexity halfway through
  • Initial assumptions prove wrong
  • Multiple approaches need evaluation
  • Problem scope emerges as you think

The Process

1. Start with Loose Estimate

Begin with rough thought count (Thought 1/5), adjust as complexity emerges. Don’t overthink the total—it changes.

2. Structure Each Thought

  • Build on previous context explicitly
  • Address ONE aspect per thought
  • State assumptions/uncertainties/realizations
  • Signal what next thought tackles

3. Apply Dynamic Operations

  • Expand: More complex → increase total (1/5 becomes 1/8)
  • Contract: Simpler → decrease total (3/8 becomes 3/6)
  • Revise: New insight → mark [REVISION of Thought 2]
  • Branch: Multiple paths → explore [BRANCH A] and [BRANCH B]

4. Verify Hypotheses

Use [HYPOTHESIS] for proposed solutions, [VERIFICATION] for test results. Iterate until verified. Mark final with [FINAL].

Common Use Cases

Debugging Authentication Flow

Who: Full-stack developer fixing login issues

"Users report login works initially but fails after 24 hours. JWT tokens are configured with 24h expiry. Refresh token logic exists in the backend. Help me debug why authentication breaks exactly at the 24h mark."

Architecture Decision for State Management

Who: React developer choosing state solution

"I need state management for a dashboard app with real-time data sync, optimistic updates, and offline support. Evaluate Redux Toolkit vs TanStack Query + Zustand, considering our team hasn't used either heavily."

API Design Review

Who: Backend engineer designing endpoints

"Reviewing our new REST API for user profiles. Current design has /users/:id for basic info and /users/:id/details for extended data. Does this split make sense or should we consolidate? Consider performance and frontend DX."

Performance Investigation

Who: Frontend dev solving slow renders

"React app has slow dashboard rendering. Profiler shows ProductList component re-renders 10+ times per page load. It receives products array from Redux store. Walk me through identifying the root cause and fix."

Database Schema Refactoring

Who: Engineer migrating data model

"Need to refactor our e-commerce schema. Current design has separate tables for physical/digital products with 60% duplicate columns. Evaluate single-table with discriminator vs polymorphic associations vs current approach."

Pro Tips

Not activating? Say: “Use sequential-thinking skill to analyze this step-by-step with explicit thought markers.”

Modes:

  • Explicit: Use visible Thought 1/5 markers when complexity warrants or requested
  • Implicit: Apply methodology internally for routine analysis

Revision example:

Thought 5/8 [REVISION of Thought 2]: Corrected understanding
- Original: localStorage is sufficient for tokens
- Why revised: XSS vulnerability discovered in dependencies
- Impact: Must switch to httpOnly cookies

Branching example:

Thought 4/7 [BRANCH A from Thought 2]: Redux Toolkit
Pros: Mature, predictable. Cons: Boilerplate, learning curve

Thought 4/7 [BRANCH B from Thought 2]: Zustand
Pros: Simple API, TypeScript. Cons: Less middleware ecosystem

Key Takeaway

Sequential Thinking transforms messy analysis into structured thought chains with revision capability, branching for alternatives, and hypothesis verification—use explicitly for complex problems or implicitly when structured reasoning improves accuracy.