perf-analyzer

The perf-analyzer is a specialized skill within the AgentSys ecosystem designed to transform raw performance investigation results into structured, evidence-based recommendations. It is intended for software engineers, performance testers, and system architects who need to parse complex profiling data, experimental results, and hypothesis outcomes without getting lost in noise. By standardizing the synthesis process, it ensures that performance tuning efforts are rooted in actual logs and code evidence rather than conjecture.

• Automatically synthesizes baseline data, profiling metrics, and breaking point results.

• Generates structured, readable output including executive summaries, actionable recommendations, and identified abandoned hypotheses.

• Enforces strict evidence requirements, mandating that all advice be directly traceable to provided logs or code snippets.

• Reduces manual overhead in performance audits by distilling large datasets into clear, next-step directives.

• Improves project consistency by adhering to standard performance documentation contracts.

• Utilize this skill when you have completed a performance investigation and need to formulate a remediation plan.

• Ensure you have your baseline data, profiling evidence, and tested hypotheses ready for the agent to consume as input.

• The agent is designed to be rigorous; if provided data is insufficient to support a conclusion, it will explicitly request a re-run of experiments rather than hallucinating performance fixes.

• Use as a phase-gated tool in your development pipeline to gate performance improvements before merging changes.

• Follow the canonical perf-requirements.md documentation for advanced integration patterns within the AgentSys framework.