Diagnose and Resolve Technical Issues Systematically

The Problem Solver Specialist is an autonomous agent that systematically investigates complex technical challenges by gathering evidence from multiple sources, analyzing patterns, and providing actionable solutions with clear reasoning. It works in five stages: first parsing the reported issue to separate core symptoms from secondary effects, classifying the problem type (performance, functionality, integration, configuration), and listing initial hypotheses; then gathering evidence by searching the codebase with Glob, examining logs, configuration files, and recent changes with Read, finding patterns and error messages with Grep, running diagnostic commands via Bash, and researching known issues with WebSearch; then correlating findings across those sources, testing each hypothesis against the evidence, and tracing the problem chain from symptom back to origin; then designing targeted fixes for the identified root cause, weighing side effects and dependencies, and prioritizing solutions by impact versus implementation complexity; and finally building a validation plan with reproduction test scenarios, success criteria, and monitoring points to prevent regression.

Use it for complex technical issues where the root cause is not obvious and needs correlation across multiple evidence sources - logs, code, config, and external research together, not a single-source guess. It is well suited to production incidents, integration failures, or intermittent bugs where the investigation itself is the hard part. It is not meant for problems with a known, obvious fix that don't need a multi-source investigation - its process is deliberately thorough (evidence gathering, hypothesis testing, risk assessment, rollback planning), which is more overhead than a one-line fix needs.

Given a reported issue, it outputs a Problem Summary (concise issue statement, classification with severity, affected scope), Investigation Results (evidence found per source, the primary root cause with supporting evidence, and contributing factors), a Recommended Solution with specific implementation steps and code examples where applicable, a Verification Plan (test steps, success criteria, monitoring recommendations), and a Risk Assessment covering implementation risk level, potential side effects with mitigations, and a rollback plan. Every report concludes with an explicit High/Medium/Low confidence level for the root cause analysis and solution recommendation, along with the factors driving that confidence.

Engineers and on-call responders facing a technical problem whose cause isn't obvious and who need a documented, evidence-based investigation trail rather than a guess-and-check fix. It suits teams that want every conclusion tied to concrete evidence and every fix accompanied by a rollback plan and regression-monitoring points, and that value a stated confidence level so the team knows how much to trust the diagnosis before acting on it.