performance-analysis

---

LIBRARY-FIRST PROTOCOL (MANDATORY)

Before writing ANY code, you MUST check:

Step 1: Library Catalog

• Location: .claude/library/catalog.json
• If match >70%: REUSE or ADAPT

Step 2: Patterns Guide

• Location: .claude/docs/inventories/LIBRARY-PATTERNS-GUIDE.md
• If pattern exists: FOLLOW documented approach

Step 3: Existing Projects

• Location: D:\Projects\*
• If found: EXTRACT and adapt

Decision Matrix

Match	Action
Library >90%	REUSE directly
Library 70-90%	ADAPT minimally
Pattern exists	FOLLOW pattern
In project	EXTRACT
No match	BUILD (add to library after)

---

STANDARD OPERATING PROCEDURE

Purpose

Investigate performance regressions or capacity risks, quantify impact, and propose validated fixes with experiments.

Trigger Conditions

• Positive: Performance regression or latency spike; Capacity planning or throughput tuning; Evidence-based performance review
• Negative: Need profiling session only (route to when-profiling-performance-use-performance-profiler); Infra build or rollout (route to infrastructure/cloud-platforms); Pure observability setup (route to opentelemetry-observability)

Guardrails

• Structure-first: keep SKILL.md aligned with examples/, tests/, and any resources/references so downstream agents always have scaffolding.
• Adversarial validation is mandatory: cover boundary cases, failure paths, and rollback drills before declaring the SOP complete.
• Prompt hygiene: separate hard vs. soft vs. inferred constraints and confirm inferred constraints before acting.
• Explicit confidence ceilings: format as 'Confidence: X.XX (ceiling: TYPE Y.YY)' and never exceed the ceiling for the claim type.
• MCP traceability: tag sessions WHO=operations-{name}-{session_id}, WHY=skill-execution, and capture evidence links in outputs.
• Avoid anti-patterns: undocumented changes, missing rollback paths, skipped tests, or unbounded automation without approvals.

Required Artifacts

• SKILL.md (this SOP)
• examples/ for analysis patterns
• tests/ or scripts in _shared/ for validation
• resources/ for dashboards

Execution Phases

1. Frame the hypothesis

   • Capture symptoms, SLO/SLA impact, and success criteria
   • List signals to collect and separate hard/soft constraints
   • Establish baseline and target thresholds

2. Collect evidence

   • Gather traces/profiles/metrics and reproduction steps
   • Correlate load, errors, and resource usage
   • Identify hotspots and suspect components

3. Analyze and propose fixes

   • Validate findings across layers (app, DB, network)
   • Draft options with impact, risk, and effort
   • Plan experiments or benchmarks to confirm

4. Validate improvements

   • Run experiments and compare to baseline
   • Update dashboards/alerts to lock in gains
   • Document recommendations and next steps

Output Format

• Performance findings summary with quantified impact
• Hotspot list with evidence links
• Remediation options with impact/risk/effort trade-offs
• Experiment or benchmark plan and results
• Recommendations with owners and timeline

Validation Checklist

• Baseline and target metrics defined and measured
• Reproduction steps or datasets captured
• Experiments show measurable deltas with evidence
• Risks and rollbacks identified for proposed changes
• Confidence ceiling stated for conclusions

Confidence: 0.70 (ceiling: inference 0.70) - Performance analysis SOP centers evidence and explicit ceilings