sequential-thinking-default

name	sequential-thinking-default
description	Use structured sequential thinking for multi-step debugging, architecture decisions, or ambiguous problem framing.

Structured Sequential Thinking

When to use this skill

Use this skill when you need to:

• Debug failures where the root cause isn't obvious
• Compare multiple valid implementation approaches
• Plan multi-file changes without missing consumers
• Make architectural decisions with competing tradeoffs
• Explore solution spaces before committing to implementation

Core Protocol

See .kilocode/rules/general-workflow.md for the full protocol. Key rules:

1. Granularity: One process_thought call = one reasoning step, ≤4 sentences
2. Stage purity: stage field must match the thought's content
3. Stage coverage: ≥3 distinct stages before entering Conclusion
4. Session hygiene: clear_history → think → generate_summary → export_session
5. Metadata required: At least one of tags, axioms_used, or assumptions_challenged per thought

Stage Progression

1. Problem Definition (≥1 thought) — What are we solving?
2. Research (≥1 thought) — What information do we need?
3. Analysis (≥2 thoughts) — Compare approaches, evaluate tradeoffs
4. Synthesis (≥1 thought) — How do pieces connect? Call generate_summary here.
5. Conclusion (≥1 thought) — Decision with rationale

Minimum: 6 thoughts across ≥3 stages for non-trivial tasks.

Thinking Styles (Composable Contracts)

For specialized reasoning, use a thinking-style contract:

Style	Contract	When to Use
Abstract	abstract_thinking.md	Generate competing problem frames
Adversarial	adversarial_thinking.md	Falsify plans, enumerate failure modes
Systems	systems_thinking.md	Find feedback loops, bottlenecks, leverage
Concrete	concrete_thinking.md	Collapse ambiguity into executable steps
Epistemic	epistemic_thinking.md	Separate know/believe/guess, assign confidence

Handoff format: thinking_handoff.md

Thinking Plans (Delegation Recipes)

For multi-style reasoning, orchestrators use thinking plan templates:

Plan	Sequence	When to Use
design-subsystem.md	Abstract → Systems → Adversarial → Concrete	Design new components
debug-incident.md	Concrete → Research → Adversarial → Concrete	Debug production issues
evaluate-dependency.md	Abstract → Systems → Adversarial → Concrete	Evaluate risky dependencies
strategic-decision.md	Abstract → Epistemic → Adversarial → Concrete	Sprint planning, strategic choices

Tools

• process_thought — Record stage-based thoughts with metadata
• generate_summary — Retrieve session overview (mandatory before Conclusion)
• export_session — Save session for future reference (mandatory at session end)
• import_session — Load previous session to resume work
• clear_history — Clear thought history (mandatory at new session start)

Integration with other tools

• Pair with codebase-retrieval: Use Research stage to plan information gathering, then execute retrieval
• Complement with Context7: When assumptions involve library APIs, verify with up-to-date docs
• Before major refactors: Use Analysis/Synthesis stages to evaluate approaches

Anti-Pattern: The Big Blob

Do NOT dump all reasoning into one massive thought. Each thought = one discrete step. See general-workflow.md for correct vs incorrect examples.