X

xcode-build-benchmark

by @avdleev
4.3(20)

これは、Xcodeビルドパフォーマンス最適化に特化したオープンソースのAIエージェントスキルです。クリーンビルド、インクリメンタルビルド、コンパイルホットスポット、プロジェクト設定、Swift Package Managerのオーバーヘッドをベンチマークし分析します。ビルド設定、スクリプトフェーズ、コンパイラフラグ内の「簡単な改善点」を特定し、最適化計画を生成し、承認された変更を自動的に適用することで、開発効率を大幅に向上させます。

xcodebuild-optimizationios-developmentperformanceai-agentGitHub
インストール方法
npx skills add avdlee/xcode-build-optimization-agent-skill --skill xcode-build-benchmark
compare_arrows

Before / After 効果比較

1
使用前

Xcodeプロジェクトのインクリメンタルビルドは通常30秒かかり、そのうち5秒は最適化可能で、開発効率が低下していました。

使用後

AIエージェントがビルド設定、スクリプト、コンパイラフラグを特定し最適化することで、インクリメンタルビルド時間が25秒に短縮され、開発効率が大幅に向上しました。

SKILL.md

Xcode Build Optimization Agent Skills

Open-source Agent Skills for benchmarking and optimizing Xcode build performance across clean builds, incremental builds, compile hotspots, project settings, and Swift Package Manager overhead.

Quick Start

Install all six skills (the orchestrator needs the specialist skills to work):

npx skills add https://github.com/avdlee/xcode-build-optimization-agent-skill

Then open your Xcode project in your AI coding tool and say:

Use the /xcode-build-orchestrator skill to analyze build performance and come up with a plan for improvements.

The agent will benchmark your clean and incremental builds, audit build settings, find compile hotspots, and produce an optimization plan at .build-benchmark/optimization-plan.md. No project files are modified until you explicitly approve changes.

For long-term monitoring across days, machines, Xcode versions, and teams, use RocketSim Build Insights and Team Build Insights.

Every Second Counts

A 1-second improvement on a 30-second incremental build sounds small. At 50 builds a day, that adds up to 3.5 hours per developer per year -- or 35 hours across a team of ten.

Most projects have several seconds of easy wins hiding in build settings, script phases, and compiler flags. This skill finds them.

How It Works

The orchestrator coordinates five specialist skills in a recommend-first workflow. Nothing is modified until you approve.

flowchart LR
    Orchestrator --> Benchmark["Benchmark\n(clean + incremental)"]
    Benchmark --> Compilation["Compilation\nAnalyzer"]
    Benchmark --> Project["Project\nAnalyzer"]
    Benchmark --> SPM["SPM\nAnalyzer"]
    Compilation --> Plan["Optimization\nPlan"]
    Project --> Plan
    SPM --> Plan
    Plan --> You{{"You review\n& approve"}}
    You --> Fixer["Build\nFixer"]
    Fixer --> ReBenchmark["Re-benchmark\n& verify"]

Phase 1 -- Analyze. The orchestrator benchmarks your project, runs the three specialist analyzers, and produces a prioritized optimization plan at .build-benchmark/optimization-plan.md. No project files are modified.

Use the Xcode build orchestrator to analyze build performance and come up with a plan for improvements.

Phase 2 -- Fix. Review the plan, check the approval boxes for the items you want, and ask the agent to apply them. The fixer implements only approved changes and re-benchmarks to verify.

Implement the approved items from the optimization plan at .build-benchmark/optimization-plan.md, then re-benchmark to verify the improvements.

The plan file is your evidence trail -- shareable with teammates, reviewable in PRs, and diffable over time.

What It Checks

The agent runs over 40 individual checks across build settings, project configuration, source code, and package dependencies.

CheckWhat the agent looks for
Build settings auditDebug/Release/General settings against best practices (compilation mode, optimization level, eager linking, compilation caching)Details
Script phase analysisMissing input/output declarations, scripts running unnecessarily, debug/simulator guardsDetails
Compile hotspot detectionLong type-checks, complex expressions, compiler diagnostic flagsDetails
Zero-change build overheadFixed-cost phases (codesign, validation, scripts) inflating incremental buildsDetails
Target dependency reviewAccuracy, parallelism blockers, monolithic targetsDetails
Module variant detectionConfig drift across targets causing duplicate module buildsDetails
SPM graph analysisPlugin overhead, branch pins, package layering, circular dependenciesDetails
Swift macro impactCascading rebuilds, swift-syntax universal buildsDetails
SwiftUI view decompositionMonolithic body properties, result builder complexityDetails
Asset catalog parallelismSingle-threaded compilation bottleneck, splitting for parallel buildsDetails
Access control optimizationMissing final, overly broad visibility inflating compiler workDetails
Incremental build diagnosticsPlanning Swift module, SwiftEmitModule, Task BacktracesDetails

Community Results

Real-world improvements reported by developers who used these skills. Add your own by opening a pull request.

The xcode-build-orchestrator generates your table row at the end of every optimization run, so contributing is a single copy-paste.

[!NOTE] A small clean-build increase is normal when enabling compilation caching -- the first cold build populates the cache. Cached clean builds (branch switching, pulling changes) and incremental builds are where the real gains show up.

AppClean BuildIncremental Build
Helm for App Store Connect86s → 91s (+5s / within noise)70s → 9s (-61s / 87% faster)
Stock Analyzer41.5s → 33.2s (-8.3s / 20% faster)5.3s → 3.6s (-1.7s / 32% faster)
Enchanted19.4s → 16.6s (-2.8s / 14% faster)2.5s → 2.2s (-0.3s / 12% faster)
Wikipedia iOS48.7s → 46.5s (-2.2s / 5% faster)12.9s → 12.2s (-0.7s / 5% faster)
Kickstarter iOS83.4s → 83.5s (~0s / within noise)10.9s → 10.6s (-0.3s / 3% faster)
Dash67.8s → 66.9s (-0.9s / within noise)6.7s → 7.1s (+0.4s / within noise)
Klivvr247ss → 167s (-80s / 32% faster)48s → 48s (same)

Who This Is For

  • iOS and macOS teams with slow local build loops
  • developers investigating a recent build-time regression
  • teams that want evidence-backed Xcode build optimization instead of guesswork
  • developers who want a reusable Agent Skills package, not a one-off script

Included Skills

SkillPurpose
xcode-build-orchestratorEnd-to-end workflow: benchmark, analyze, prioritize, approve, fix, re-benchmark
xcode-build-benchmarkRepeatable clean and incremental build benchmarks with timestamped artifacts
xcode-compilation-analyzerSwift compile hotspot analysis and source-level recommendations
xcode-project-analyzerBuild settings, scheme, script phase, and target dependency auditing
spm-build-analysisPackage graph, plugin overhead, and module variant review
xcode-build-fixerApply approved optimization changes and verify with benchmarks

The orchestrator is the recommended starting point -- it coordinates the other five skills automatically. Install all six skills so the orchestrator can access each specialist.

Installation Options

Option A: Using skills.sh

Install all six skills (required -- the orchestrator depends on the specialist skills):

npx skills add https://github.com/avdlee/xcode-build-optimization-agent-skill

To install a single skill for standalone use, add the --skill flag:

npx skills add https://github.com/avdlee/xcode-build-optimization-agent-skill --skill xcode-project-analyzer

Available individual skills: xcode-build-benchmark, xcode-compilation-analyzer, xcode-project-analyzer, spm-build-analysis, xcode-build-orchestrator, xcode-build-fixer. Note that the orchestrator requires all other skills to be installed.

Option B: Claude Code Plugin

  1. Add the marketplace:
/plugin marketplace add AvdLee/Xcode-Build-Optimization-Agent-Skill
  1. Install the plugin:
/plugin install xcode-build-skills@xcode-build-skills

To enable for everyone in a repository, add to your project configuration:

{
  "enabledPlugins": {
    "xcode-build-skills@xcode-build-skills": true
  },
  "extraKnownMarketplaces": {
    "xcode-build-skills": {
      "source": {
        "source": "github",
        "repo": "AvdLee/Xcode-Build-Optimization-Agent-Skill"
      }
    }
  }
}

Option C: Cursor Plugin (coming soon)

This repository is packaged for Cursor plugin submission, but the marketplace listing is not live yet.

Once approved, you'll be able to install it from the Cursor Marketplace.

Option D: Codex / OpenAI-compatible tools

This repository includes an agents/openai.yaml manifest. Copy or symlink the skill folders into your Codex skills directory:

cp -R skills/ "$CODEX_HOME/skills/"

See Codex skills documentation for details on where to save skills.

Option E: Using pi package manager

Install via pi:

pi install https://github.com/AvdLee/Xcode-Build-Optimization-Agent-Skill

The skills will be available automatically in pi sessions.

Option F: Manual Install

  1. Clone this repository.
  2. Install or symlink the specific skill folder from skills/ that you want.
  3. Ask your AI coding tool to use the corresponding skill.

Useful docs: Codex Skills | Claude Code Agent Skills | Cursor Skills

Why Clean And Incremental Builds Both Matter

Clean builds expose:

  • package and module setup cost
  • full project graph overhead
  • target structure and explicit-module issues

Incremental builds expose:

  • edit-loop pain
  • run script bottlenecks
  • cache invalidation problems
  • repeated package-plugin overhead

That distinction is central to this repo and follows both Apple's Xcode guidance and the SwiftLee workflow in Build performance analysis for speeding up Xcode builds.

See Also My Other Skills

Shared Support Layer

Each skill bundles its own copies of the scripts, references, and schemas it needs so it works after standalone installation. The canonical copies live at the repo root:

  • scripts/ -- helper scripts for benchmarking, timing-summary parsing, compilation diagnostics, report generation, and recommendation rendering
  • references/ -- build settings best practices, artifact format, recommendation format, and source citations
  • schemas/ -- JSON schema for benchmark output

When a root-level file changes, the corresponding copies inside each skill that uses it must be updated (see CONTRIBUTING.md).

Skill Structure

skills/
  xcode-build-benchmark/
    SKILL.md
    references/
      benchmark-artifacts.md
      benchmarking-workflow.md
  xcode-compilation-analyzer/
    SKILL.md
    references/
      build-optimization-sources.md
      code-compilation-checks.md
      recommendation-format.md
  xcode-project-analyzer/
    SKILL.md
    references/
      build-optimization-sources.md
      build-settings-best-practices.md
      project-audit-checks.md
      recommendation-format.md
  spm-build-analysis/
    SKILL.md
    references/
      build-optimization-sources.md
      recommendation-format.md
      spm-analysis-checks.md
  xcode-build-orchestrator/
    SKILL.md
    references/
      benchmark-artifacts.md
      build-settings-best-practices.md
      orchestration-report-template.md
      recommendation-format.md
  xcode-build-fixer/
    SKILL.md
    references/
      build-settings-best-practices.md
      fix-patterns.md
      recommendation-format.md

Research Basis

All checks are grounded in Apple documentation, WWDC sessions, and proven community practices. See OPTIMIZATION-CHECKS.md for the full list of checks with references to each source.

The stored reference summaries live in references/build-optimization-sources.md.

RocketSim Positioning

This repo helps you optimize point-in-time build performance with an agent-guided workflow.

RocketSim complements it by monitoring build performance over time:

  • automatic clean vs incremental build tracking
  • duration trends and percentile metrics
  • machine, Xcode, and macOS comparisons
  • team-wide visibility without custom build scripts

If you want to catch regressions earlier and see whether your build times are improving over weeks or months, use RocketSim Build Insights after you apply the improvements from this repo.

Contributing

Contributions are welcome when they keep the repo focused on Xcode build optimization and Agent Skills format quality.

Please read CONTRIBUTING.md for:

  • skill-authoring guidance
  • repo scope and quality standards
  • workflow notes for scripts and README sync

About The Author

Created by Antoine van der Lee, creator of SwiftLee and RocketSim. The practical build workflow in this repository is informed by the SwiftLee article Build performance analysis for speeding up Xcode builds and ongoing work on RocketSim Build Insights.

License

This repository is available under the MIT License. See LICENSE for details.

ユーザーレビュー (0)

レビューを書く

効果
使いやすさ
ドキュメント
互換性

レビューなし

統計データ

インストール数2.1K
評価4.3 / 5.0
バージョン
更新日2026年5月20日
比較事例1 件

ユーザー評価

4.3(20)
5
60%
4
40%
3
0%
2
0%
1
0%

この Skill を評価

0.0

対応プラットフォーム

🔧Claude Code

タイムライン

作成2026年3月31日
最終更新2026年5月20日