code-review-performance

スター7

フォーク0

更新日2026年2月17日 08:24

Run performance-focused code review when changes may affect latency, throughput, or CPU/memory/I/O efficiency on critical paths. Use for merge decisions requiring explicit performance-risk findings; do not use for broad non-performance review scope.

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

ソース

KentoShimizu

KentoShimizu/sw-agent-skills

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Code Review Performance

Overview

Use this skill to detect performance regressions before merge, especially on hot paths and high-traffic execution flows.

Scope Boundaries

Use this skill when the task matches the trigger condition described in description.
Do not use this skill when the primary task falls outside this skill's domain.

Inputs To Gather

Hot-path endpoints/jobs and current performance budgets.
Workload assumptions (QPS, payload size, concurrency, data cardinality).
Existing benchmark/profiling evidence.
Resource constraints (CPU, memory, I/O, network).

Deliverables

Performance findings prioritized by user impact.
Budget-fit judgment (within budget / at risk / out of budget).
Required follow-up checks (benchmark, profiling, load test).

Finding Focus Areas

Algorithmic growth (O(n^2) regressions, repeated scans).
Allocation pressure and unnecessary object churn.
I/O amplification (N+1 calls, repeated DB/API access).
Contention/serialization bottlenecks under concurrency.
Cache invalidation or cache-bypass risks.

Quick Example

Change adds per-item DB call inside loop over 10k records.
Finding: high-severity throughput risk (N+1 query pattern).
Fix direction: batch query + in-memory map, verify via benchmark.

Quality Standard

Each finding ties to an explicit performance budget or hotspot.
Recommendations include measurement plan, not assumptions only.
Risk classification includes expected scale sensitivity.
Missing evidence (benchmark/profile) is flagged explicitly.

Workflow

Identify changed hot paths and performance-sensitive flows.
Analyze algorithmic and resource behavior from diff.
Compare expected behavior with existing budgets.
Require measurement where uncertainty is material.
Publish findings with mitigation and verification steps.

Failure Conditions

Stop when high-risk regressions have no mitigation/verification plan.
Escalate when performance impact cannot be bounded from available evidence.

このリポジトリの他の Skills

同じリポジトリ

debug-root-cause

KentoShimizu/sw-agent-skills

Systematic debugging workflow for isolating software root causes, implementing proportional fixes, and verifying recovery with reproducible evidence. Use when application/runtime failures, regressions, or flaky behavior are observed and the task requires both diagnosis and remediation. When specialized skills are available in the environment, defer GitHub Actions failure triage to `github-fix-ci`, commit-introduction isolation to `git-bisect-debugging`, profiler-led hotspot analysis to `performance-profiling`, active security incident command to `security-incident-response`, and postmortem-only reporting to `incident-postmortem`. Do not use for broad feature implementation or speculative hardening without evidence.

2026-03-027

api-design-rest

KentoShimizu/sw-agent-skills

Resource-oriented REST/OpenAPI contract design for URI/method semantics, idempotency, pagination, and HTTP status behavior. Trigger when REST contract diffs are detected in paths, methods, schemas, or status semantics, or when request-response transport choice is still unresolved before implementation. Do not use for GraphQL schema authoring, error-taxonomy-only changes, version lifecycle governance, or consumer-provider contract test implementation.

2026-02-177

ab-testing

KentoShimizu/sw-agent-skills

Controlled online experiment workflow for product changes with causal inference, randomization integrity checks, and pre-registered decision criteria. Trigger when ship/no-ship decisions require causal evidence under uncertainty and KPI trade-offs must be quantified. Do not use for feature-flag rollout policy without experiment design, deterministic functional verification, or observability-only analysis.

2026-02-177

testing-contract

KentoShimizu/sw-agent-skills

Provider-consumer compatibility testing for service interface changes. Use when APIs or event schemas evolve and executable contract checks must guard compatibility before release; do not use for UI-only validation or architecture topology decisions.

2026-02-177

testing-e2e

KentoShimizu/sw-agent-skills

End-to-end test planning for critical user journeys across integrated systems. Use when release confidence depends on browser/API/dependency behavior as one flow; do not use for isolated unit behavior or pure contract checks.

2026-02-177

testing-regression-strategy

KentoShimizu/sw-agent-skills

Risk-based regression suite curation for release gating under time and budget limits. Use when teams must decide what runs always versus conditionally by impact and risk; do not use for implementing a single test type in isolation.

2026-02-177

Code Review Performance

Overview

Use this skill to detect performance regressions before merge, especially on hot paths and high-traffic execution flows.

Scope Boundaries

Use this skill when the task matches the trigger condition described in description.

Do not use this skill when the primary task falls outside this skill's domain.

Inputs To Gather

Hot-path endpoints/jobs and current performance budgets.

Workload assumptions (QPS, payload size, concurrency, data cardinality).

Existing benchmark/profiling evidence.

Resource constraints (CPU, memory, I/O, network).

Deliverables

Performance findings prioritized by user impact.

Budget-fit judgment (within budget / at risk / out of budget).

Required follow-up checks (benchmark, profiling, load test).

Finding Focus Areas

Algorithmic growth (O(n^2) regressions, repeated scans).

Allocation pressure and unnecessary object churn.

I/O amplification (N+1 calls, repeated DB/API access).

Contention/serialization bottlenecks under concurrency.

Cache invalidation or cache-bypass risks.

Quick Example

Change adds per-item DB call inside loop over 10k records.

Finding: high-severity throughput risk (N+1 query pattern).

Fix direction: batch query + in-memory map, verify via benchmark.

Quality Standard

Each finding ties to an explicit performance budget or hotspot.

Recommendations include measurement plan, not assumptions only.

Risk classification includes expected scale sensitivity.

Missing evidence (benchmark/profile) is flagged explicitly.

Workflow

Identify changed hot paths and performance-sensitive flows.

Analyze algorithmic and resource behavior from diff.

Compare expected behavior with existing budgets.

Require measurement where uncertainty is material.

Publish findings with mitigation and verification steps.

Failure Conditions

Stop when high-risk regressions have no mitigation/verification plan.

Escalate when performance impact cannot be bounded from available evidence.

name	code-review-performance
description	Run performance-focused code review when changes may affect latency, throughput, or CPU/memory/I/O efficiency on critical paths. Use for merge decisions requiring explicit performance-risk findings; do not use for broad non-performance review scope.