sequential-think_skill

Overview

This skill is a multi-step reasoning engine that guides systematic, iterative analysis for complex problems. It is designed for layered debugging, architecture evaluation, hypothesis-driven investigation, and performance bottleneck identification. Use it when a problem requires several coordinated thinking steps rather than a single-shot answer.

How this skill works

Each invocation represents one explicit "thought" in a chain: identify scope, break into subproblems, generate hypotheses, verify evidence, iterate, and conclude. Run via an available MCP tool or a standalone CLI script to append, revise, branch, or finalize thoughts and to inspect thought history. Outputs include metadata (thoughtNumber, totalThoughts, nextThoughtNeeded, branches) so analysis can be audited and continued.

When to use it

• Complex debugging across multiple layers or services
• System architecture analysis and design trade-offs
• Investigations that require hypothesis testing and verification
• Multi-component failure analysis and root-cause discovery
• Performance bottleneck identification and progressive narrowing

Best practices

• Start with an estimated totalThoughts and adjust as understanding evolves
• Mark revisions explicitly (isRevision + revisesThought) to keep history clear
• Branch for alternative approaches using branchFromThought and meaningful branchId
• Keep each thought focused: advance the analysis or test a hypothesis
• Only set nextThoughtNeeded=false after verifying conclusions against evidence

Example use cases

• Triage a production outage involving network, database, and application layers
• Design a scalable service by decomposing components and testing capacity hypotheses
• Investigate intermittent performance regressions with stepwise narrowing
• Explore multiple remediation strategies in separate branches before committing
• Document an audit trail of reasoning for peer review or handoff