Overview

The perf-theory-gatherer crafts testable performance hypotheses for a defined scenario by inspecting recent Git history and identifying relevant code paths. It anchors each hypothesis in concrete evidence and assigns a confidence level, following the canonical contract in docs/perf-requirements.md.

How This Skill Works

It reviews recent git history scoped to relevant paths when possible, uses repo-map or grep to identify the code paths involved in the scenario, and outputs up to 5 hypotheses with an id, short description, evidence, and a confidence label.

When to Use It

• Before optimizing a feature, to prioritize hotspots based on evidence
• After a performance regression to quickly locate affected components
• When evaluating candidate optimizations, anchored to recent changes
• During performance investigations of critical paths identified in the repo
• When documenting performance hypotheses for a PR or design review

Quick Start

1. Step 1: Review recent git history for the scenario scope
2. Step 2: Identify code paths involved using repo-map or grep
3. Step 3: Produce up to 5 hypotheses with evidence and confidence in the required format

Best Practices

• Limit git history review to paths relevant to the scenario
• Review the most recent changes first; avoid broad history
• Use repo-map or grep to identify code paths involved
• Output exactly up to 5 hypotheses with evidence and confidence
• Do not propose optimizations; focus on hypotheses and evidence only

Example Use Cases

• H1: Latency spike in GET /users after recent auth middleware change; evidence: src/middleware/auth.rs, commit 1234abcd; confidence: high
• H2: Throughput drop in data-ingest job after loop optimization; evidence: src/processors/ingest.go lines 210-260; confidence: medium
• H3: UI render delay following React component refactor; evidence: src/ui/components/ItemList.jsx change set; confidence: medium
• H4: Database query path slow after ORM query modification; evidence: orm/query.go, commit 5f9a2e; confidence: high
• H5: External API call becomes bottleneck after new retry logic; evidence: services/external.go, retryPolicy.go; confidence: low

Frequently Asked Questions