Guides, comparisons, and best practices for AI coding rules. Learn how to write effective CLAUDE.md, .cursorrules, and AGENTS.md files for your projects.
A local copied-secret experiment separates vault access removal, credential rotation, process restart, and audit needs. Use a requirements checklist to compare manual handling with managed secrets without assuming a paid vault revokes exported credentials.
A three-case SDK 1.27.1 reproduction: initialize and tool calls can pass while a stray stdout log produces a parser error. Test channel cleanliness separately from functional success.
A reproducible TypeScript SDK 1.27.1 stdio test: distinguish Node inheritance from SDK defaults and pass only the variables your server needs.
Separate agent instructions, ownership review, and merge controls. A practical preflight for sensitive repository changes.
v2.1.283 stopped cmd /c rd, rmdir, del, and erase from deleting drive roots on Windows. On Sept 27, a bash subagent routed the same class of deletion through MSYS2 instead and wiped 125GB.
availableModels never pinned versions the way most teams assumed. Claude Code v2.1.283 added deniedModels and availableModelsMatch to actually close the gap — here's the precedence and config templates.
v2.1.283 gave /doctor a prompt-audit subcommand for CLAUDE.md, skills, agents and commands — not the same feature as the older /claude-api prompt-audit. Here's what each one actually checks.
Six permission-check and managed-settings bugs fixed in Claude Code v2.1.282: CLAUDE.md symlink escapes, skipped Bash rules, duplicate remote runs, mistyped locks.
v2.1.281 and v2.1.282 patched five separate claude plugin uninstall bugs on Sept 23-24. Nine open GitHub issues show the fix doesn't reach the Desktop app or claude.ai sync layer.
Three GitHub issues filed within 5 days of Claude Code v2.1.277 show AGENTS.md fallback loading depends on a live feature-flag fetch, not just files on disk.
v2.1.280 lets you raise Claude Code's 2,048-character MCP description cap. Three closed GitHub issues asked for four fixes over six months — this ships one, and the new variable isn't in the docs yet.
APM, Microsoft's package manager for AI agent config, writes zero instruction files to Cursor's user scope — silently. A same-day GitHub issue explains why, and why the fix stalled.
v2.1.278 made auto mode's classifier checks free — then broke 4 community gateways within days. What the fix requires, and the doc gap causing it.
v2.1.277 didn't just add an AGENTS.md fallback — it added a Project instructions setting with four values, a CLAUDE.local.md gotcha, and per-subdirectory rules almost no launch-week coverage mentioned.
A skill declaring 5 tools in allowed-tools can receive 46+, undeclared write access included. A separate open bug means fenced code blocks documenting the ! syntax execute for real. What v2.1.271 changed, and how to actually contain a skill.
claude plugin eval runs a plugin's eval suite against a no-plugin baseline and scores the Δ. The official docs say the early-access error just means 'update your CLI.' One team's GitHub issue shows that alone didn't work — here's the full picture, sourced from the spec and the paper trail.
CLAUDE_CODE_GATEWAY_HINT_HEADERS=1 makes Claude Code send request-class, agent-type, and three more headers to LLM gateways. Here's what each one is for, and how to use them before the docs catch up.
Claude Code v2.1.271 added omitClaudeMd to subagent frontmatter, so self-contained subagents can launch without CLAUDE.md. What it does, the managed-policy exception, and when it beats the directory-split workaround.
CVE-2026-54316 let a pre-approved huggingface.co domain leak data from Claude Code one character at a time. How it worked, why its CVSS scores disagree, and the settings.json pattern worth adding today.
Issue #6235 asked Claude Code to read AGENTS.md natively. It got 5,146 👍, then an auto-close bot marked it 'completed.' What @import and /import actually do in 2026.
Effect v4 hit Release Candidate on August 12, 2026, and npm's own dist-tags show it's still not the default install. Context.Tag, Context.GenericTag, Effect.Tag, and Effect.Service all collapse into Context.Service — verified against the official effect-smol migration docs, not a blog summary.
Since Playwright switched to Chrome for Testing, npx playwright install chromium downloads a full browser plus a separate headless shell — roughly 300MB — unless you pass --only-shell. Verified against playwright-core's own browsers.json.
MCP TypeScript SDK v2 split the old package into 9 modules on July 27. Six weeks later, 86% of installs are still v1.
Tabnine Agent doesn't read AGENTS.md or CLAUDE.md. It uses .tabnine/guidelines/*.md instead, and org-level Admin Console rules silently override your personal ones. Sourced from Tabnine's own docs.
pnpm 12's Rust rewrite kept the lockfile format, but git dependency resolution, engineStrict, and global bins changed. What CLAUDE.md needs to say before an agent hits any of it.
replit.md isn't a static rules file like CLAUDE.md or AGENTS.md — Replit Agent generates it, reads it, and silently edits it as it works. Here's how that changes what you should (and shouldn't) put in it.
Biome doesn't recognize eslint-disable comments at all — they just sit there as dead text while the lint rule still fires. The CLAUDE.md rules that stop your agent from writing them.
Qwen Code added automatic AGENTS.md support in March 2026 (PR #2018), loaded alongside QWEN.md — but only one of the two files gets written to. Verified against the CLI's own source, not just the docs.
Tailwind CSS v4 pulls over 125M weekly npm downloads, but its zero-config setup and silent default changes still trip up Claude Code. The exact CLAUDE.md rules that close the gap.
TanStack Start is now roughly a third of Next.js's download volume, but most CLAUDE.md files have zero rules for it. The decision framework and the AI coding rules that fix it.
The concrete Apollo Client mistakes Claude Code and similar agents ship in React apps — cache, useMutation, and Apollo Client 4 imports — with the fix for each.
Hono passed Express as the go-to choice for edge APIs in 2026. The Context API, middleware model, and RPC type inference all break in specific ways when Claude Code writes Hono like Express.
TypeScript 6.0 flipped strict, module, target, and types to new defaults and dropped ES5. What that does to the CLAUDE.md rules teams already wrote to enforce the old behavior.
On August 20, 2026, anthropic hit 1.0.0 on a new HTTP layer (httpx2) and every pydantic-ai release up to that day let it install anyway. Here's what actually broke, and the CLAUDE.md pinning rule that prevents the next one.
Three unrelated things now answer to the name Remix. What CLAUDE.md has to state so Claude Code doesn't guess wrong on imports, APIs, or the rendering model.
Bun 1.4 replaced roughly a million lines of Zig with Rust, much of it AI-assisted, and shipped four new builtin APIs in the same release. What CLAUDE.md needs to say before Claude Code reaches for bun upgrade or swaps in a new builtin.
Redis changed its license twice since 2024 and split into two projects — Redis (tri-licensed) and Valkey (BSD). What CLAUDE.md needs to state so Claude Code doesn't guess wrong on the engine, the Docker image, or the modules your commands depend on.
AWS Lambda's new Durable Functions model checkpoints your code and replays it from the top on every resume — which silently breaks if Claude Code writes a timestamp, a UUID, or an API call outside a step. The CLAUDE.md rules, decision framework, and guardrails that catch it before it ships.
A CLAUDE.md template for Drizzle ORM — which driver import matches your actual database, why db:push destroys production schema history, and the relational query API's nested-shape gotcha.
A CLAUDE.md template for LangChain apps covering the v1.0 langchain-classic package split, the three generations of agent creation, and version-pinned imports.
A CLAUDE.md template for Unity projects — pinning Editor version and render pipeline, why direct edits to .unity/.prefab/.meta files corrupt projects, and setting up MCP for Unity safely.
A CLAUDE.md template for Cloudflare's Agents SDK — setState vs plain fields, the schedule() callback-as-string trap, binding-name vs URL kebab-casing, and what survives hibernation.
The AI SDK's 'ai' package pulls 23M+ weekly npm downloads, but three major versions in under two years renamed nearly every core API. Rules to stop Claude Code writing v4-era code in a v7 project.
Axum has no dedicated CLAUDE.md guide anywhere in 2026 despite 437M+ crates.io downloads. Rules for extractor ordering, the 0.8 {id} path migration, and layer vs ServiceBuilder ordering.
A complete CLAUDE.md template for Godot 4 and GDScript — static typing, autoload managers, signal conventions, and why an undocumented project pushes Claude Code toward the deprecated TileMap node.
A dedicated CLAUDE.md template for Hotwire — Stimulus targets/values/outlets, Turbo Frame vs. Stream boundaries, and the lifecycle-ordering mistakes Claude Code makes without explicit rules.
How Claude Code works with GitLab in 2026: the GitLab CI/CD integration with @claude mentions, --worktree support for merge request URLs, gitlab.com plugin marketplaces, and token redaction.
Convex generates and updates part of your CLAUDE.md automatically via npx convex ai-files. Here's exactly what it manages, what it leaves alone, and how to write the rest.
GraphQL's flexibility is also its failure mode: an unguided resolver query is invisible until it's N+1 in production. CLAUDE.md rules that catch it first.
tRPC's whole point is that a broken contract fails to compile. A CLAUDE.md that stops agents from working around that guarantee instead of using it.
A CLAUDE.md for Turborepo that covers turbo.json task pipelines, remote caching, Composable Configuration, and the experimental boundaries command — not generic monorepo advice.
Prisma projects fail the same way across frameworks: bad migrations, connection exhaustion, N+1 queries. A copy-paste CLAUDE.md that stops all three.
A CLAUDE.md template for Cloudflare Workers — wrangler.jsonc bindings (KV, D1, R2, Durable Objects, Queues), compatibility_date discipline, how official Cloudflare Skills fit alongside CLAUDE.md, and hook-based deploy safety gates.
Set up Claude Code with Firebase: MCP server config, Agent Skills, a CLAUDE.md template for Firestore/Cloud Functions, and Emulator Suite workflows.
A complete Claude Code setup guide for React Native and Expo — production CLAUDE.md template, Expo Skills/MCP integration, EAS Build CI/CD, native module patterns, and common pitfalls.
A complete CLAUDE.md template for WordPress plugin and theme development — nonces, capability checks, and $wpdb->prepare() as non-negotiable rules, block themes vs. classic theme structure, and the new Abilities API that lets Claude Code call your site directly through MCP.
A complete CLAUDE.md template for Astro — the hydration directive decision tree, Content Layer loaders, server:defer islands, and how Astro 7's built-in AI-agent dev server changes your hooks.
A complete CLAUDE.md template for Express.js APIs — the Express 4-to-5 async error handling shift, the 4-argument error middleware trap, and why NestJS's structure doesn't transfer to plain Express.
A complete CLAUDE.md template for Kubernetes projects — Helm chart authoring, ArgoCD/Flux GitOps rules, OPA/Gatekeeper guardrails, and what real-world rule files get wrong.
A CLAUDE.md template for Svelte 5 and SvelteKit — runes over legacy reactivity, when to use remote functions vs. form actions, SvelteKit 3 preview changes, and AGENTS.md for component library monorepos.
A CLAUDE.md template for Angular 21+ projects — zoneless change detection, Signal Forms, standalone components, NgRx Signal Store vs. native signals, and Vitest test conventions.
A complete CLAUDE.md template for .NET projects — ASP.NET Core minimal APIs vs. controllers, EF Core migrations, DI service lifetimes, C# 14's field keyword, and dotnet CLI hooks.
A complete CLAUDE.md template for native Android development — Compose state hoisting, Hilt's annotation maze, KSP vs. KAPT, and why CLAUDE.md needs to scale per-module in large Gradle projects.
How to set up Claude Code for Laravel and PHP projects — a manual CLAUDE.md template, Laravel Boost's MCP server, and the guidelines/skills/project-rules system explained.
Complete CLAUDE.md template for Java + Spring Boot — constructor injection, layered architecture, Maven vs Gradle, Spring Modulith, and hook-driven verification with Claude Code.
A working CLAUDE.md template for testing with Jest, Vitest, and Playwright — mocking discipline, coverage rules, and how Playwright's new agent trio changes Claude Code test generation.
Every context, action, and keystroke rule in Claude Code's keybindings.json — plus working configs for reclaiming Ctrl+X, fixing tmux conflicts, and vim-mode gotchas.
How Claude Code's 5-hour and weekly usage limits actually work, what changed on May 6, 2026, how to read /usage and /insights, and the official tactics for not hitting the wall mid-session.
Auto mode becomes Claude Code's default on Pro, Max, and Team plans August 14, 2026. What changed, the safety data behind it, and how to configure or opt out via settings.json.
Claude Managed Agents and Claude Code are not the same product. This guide breaks down the architecture, pricing ($0.08/session-hour), and API differences, plus when to use each — with the Agent SDK as the third option.
How Claude Code output styles work: the four built-in styles, custom creation, file precedence, the 2025 deprecation reversal, and how they differ from CLAUDE.md and statusLine.
Every JSON field Claude Code sends to a statusLine script, subagentStatusLine for parallel agents, ready-to-use examples, and how it differs from output styles.
How Claude Code's cross-session messaging works: ListAgents, SendMessage, crossSessionInbound, dialogExpiry, and how it differs from Remote Control and agent teams.
How Claude Code self-hosted environments work: the environment/runner/session model, setup, git credentials, hardening, CLI flags, and how it differs from Remote Control.
Install Factory Droid CLI, write your first AGENTS.md, and wire it into droid exec for CI. Real templates, verification steps, and the mistakes that make Droid ignore your instructions.
Set up AGENTS.md in Roo Code: native loading order, the AGENT.md fallback, combining it with Custom Modes, subfolder discovery, and how to verify it's actually being read.
AGENTS.md loads on every turn; SKILL.md loads only when needed. We compare the two standards' origins, frontmatter, and when a repo needs one, the other, or both.
Auggie CLI checks CLAUDE.md before AGENTS.md by default. Exact rules precedence, hierarchy, and what a 2026 ETH Zurich study found about context files.
How Google Jules reads AGENTS.md, runs Gemini 3 Pro in a Cloud VM, handles secrets, and triggers via GitHub Actions — verified against official Jules docs.
How OpenHands reads AGENTS.md, why its own SDK docs file it under 'Legacy Context', and what replaced microagents: the SKILL.md-based skills directory.
Every documented settings.json field for Claude Code, organized by category, with the precedence order across managed, CLI, local, project, and user scopes.
JetBrains Junie moved from .junie/guidelines.md to AGENTS.md. Here's the discovery order, what still works, and why some users find Junie stops reading their old file.
Block's Goose reads both AGENTS.md and .goosehints, checked at every directory level from your working directory down, with a CONTEXT_FILE_NAMES setting that controls which filenames it looks for. Verified against Goose's own docs.
Lovable reads AGENTS.md and CLAUDE.md from your project root, alongside Workspace Knowledge, Project Knowledge, and Skills — four separate context layers with different scopes, limits, and precedence. Verified against Lovable's own docs.
Sourcegraph's Amp checks three OS-wide system paths for AGENTS.md that most tools skip, plus repo and home directories, with AGENT.md and CLAUDE.md as fallbacks.
Google Antigravity reads AGENTS.md and GEMINI.md as Rules, defines custom agents through agent.md frontmatter, and loads Skills from SKILL.md — exact file paths, the 12,000-character limit, and what's actually confirmed in Google's own docs.
AGENTS.md silently ignored in Codex, Claude Code, Cursor, or GitHub Copilot? Verification commands, real GitHub bug reports, and concrete tool-by-tool fixes for 2026.
Droid CLI searches .factory/, .agents/, and .agent/ folders up to the git root, and it accepts CLAUDE.md as a compatible filename. The exact precedence order, explained.
xAI open-sourced Grok Build on July 15, 2026 — and unlike Codex's import tool, it reads CLAUDE.md, skills, hooks, and MCP servers from an existing Claude Code repo with zero migration. Here's how it actually works.
Codex CLI v0.145 expanded /import to pull in CLAUDE.md, skills, hooks, MCP servers, subagents, and settings from Claude Code. Here's exactly what transfers, what breaks, and what has no equivalent.
MCP's 2026-07-28 spec finalized a stateless core, hardened OAuth/OIDC, and versioned Apps and Tasks extensions. Here's what actually changes for the MCP servers behind your Claude Code .mcp.json.
Claude Code stopped self-invoking /verify and /code-review in v2.1.215, then made /code-review run as a background subagent in v2.1.218. Here's the verified changelog history and what it changes about your workflow.
sandbox.filesystem.disabled drops filesystem isolation while keeping network control; strictAllowlist forces silent denies for CI. Exact syntax, defaults, and scope explained.
Claude Code capped, then disabled, then partially re-enabled nested subagent spawning between July 17-24, 2026. Here's the verified changelog timeline and what it means for orchestration.
Claude Code v2.1.212 redefined /fork: it now spawns an independent background session instead of an in-conversation subagent. The old behavior moved to /subtask.
Claude Opus 5 is now Claude Code's default Opus model. Exact fast mode pricing, how tool changes and safety fallback work mid-session, and when to use Sonnet 5 instead.
How /doctor became /checkup in Claude Code v2.1.205 — diagnosing and fixing installs, flagging unused skills/MCP servers/plugins by context cost, and trimming CLAUDE.md.
Five permission-check bypasses fixed in Claude Code v2.1.214: nested-dir allow rules, PowerShell 5.1, zsh subscripts, help/man, and 10,000+ char commands.
How Claude Code Remote Control works, how it differs from Dispatch, Channels, and Slack, the security model, the Trusted Devices beta, and enterprise admin policy controls.
How Claude Code's --teleport flag pulls a claude.ai cloud session into your terminal, why the handoff is one-way, and how it differs from Remote Control.
A technical breakdown of Claude Code v2.1.212–v2.1.216: automatic backgrounding for MCP tool calls past two minutes, the new EndConversation tool, the progress heartbeat, and the permission-check fixes that shipped alongside them — with the exact mechanics and env vars MCP server builders need.
Does Continue.dev read AGENTS.md, or only its own config.yaml and .continue/rules/ system? The verified answer, exact setup syntax, and how it compares to Cline and Cursor.
Claude Code artifacts can now call MCP connectors on every page view instead of freezing a session snapshot. A technical breakdown of the mechanism, setup steps, viewer-scoped permissions, and how it compares to static artifacts, Retool, and Streamlit.
A config-file-focused comparison of Devin AI and Claude Code: AGENTS.md support, Devin's Knowledge and Playbooks system, CLAUDE.md imports, and how each tool's autonomy model shapes what a rules file needs to do.
Kilo Code (now rebranded 'Kilo') and Trae AI (ByteDance) both ship their own project-rules conventions — .kilo/rules/ + kilo.jsonc, and .trae/rules/ with four application modes — and both also support AGENTS.md natively. A verified guide to file locations, syntax, precedence, and setup for each.
How JetBrains AI Assistant actually handles CLAUDE.md, AGENTS.md, and native Project Rules across IntelliJ IDEA, PyCharm, and WebStorm — which agent reads which file, exact setup steps, and where the three systems overlap.
How Zed's built-in AI agent actually reads AGENTS.md, CLAUDE.md, and legacy .rules files — the exact file priority order, personal vs. project instructions, and why the old Rules Library no longer exists.
AGENTS.md is the open multi-tool standard, CLAUDE.md is Claude Code's native format, and GEMINI.md is Gemini CLI's own convention — with a hierarchy and import system that differs from both. A verified, three-way comparison with tool support matrix and code examples.
How Warp's Agent Mode reads project instructions: AGENTS.md and WARP.md file support, Global vs. Project Rules, precedence order, Skills, and a step-by-step setup guide.
Three battle-tested AGENTS.md templates for startup, growth-stage, and enterprise teams — plus a required/recommended/optional tier system, change management workflow, and rollout playbook. Built from reviewing 200+ real repos across team sizes.
Seven production-tested strategies for managing Claude Code's context window — /compact timing, file references over pasting, sub-agent isolation, and session handoff patterns. Includes a cost-vs-effect matrix and context pressure signals to watch for before things go sideways.
Set up Linear's MCP server for Claude Code, assign issues to AI agents, track progress through sessions, and build a CLAUDE.md workflow that keeps your Linear board and code in sync.
Does GitHub Copilot Workspace read AGENTS.md, CLAUDE.md, or copilot-instructions.md? We tested all three formats and show the best config setup for teams using Copilot Workspace alongside other AI tools.
How to roll out Claude Code across a team: shared CLAUDE.md standards, AGENTS.md setup, permission mode policies, MCP server configuration, and onboarding checklists for developers, tech leads, and CTOs.
Practical CLAUDE.md templates for Django development — migrations append-only rule, manage.py commands, Django REST Framework setup, Celery patterns, and Claude Code Hooks for automated checks.
Which AI coding tools read AGENTS.md, which require CLAUDE.md, and how Kiro steering files change the picture. Updated July 2026 with corrected Continue.dev status, Kiro integration, and the @import pattern for using both.
Production-ready CLAUDE.md and AGENTS.md templates for Terraform, Kubernetes/Helm, and Ansible. Includes hook-based safety gates, per-environment scoping, and multi-tool DevOps workflow patterns.
Complete breakdown of Desktop-exclusive features — diff preview, PR monitoring, auto-fix CI, browser pane, scheduled tasks, and computer use — compared to CLI equivalents with a practical workflow guide.
Unit testing, integration testing, transport-specific strategies, and Claude Code hook-based health monitors for MCP servers. Covers MCP Inspector, stdio vs HTTP testing, Vitest/pytest patterns, and CI/CD integration.
When to use Claude Fable 5 and extended thinking mode in Claude Code. Covers model selection flags, effort levels, CLAUDE.md configuration, and which tasks benefit from deeper reasoning.
Complete guide to Claude Code hooks — UserPromptSubmit, PreToolUse, PostToolUse, Stop, and Notification hooks. Real examples for security gates, auto-logging, context injection, and quality checks. Covers the if field, multi-hook composition, and prompt-based hooks that most guides miss.
How Windsurf reads AGENTS.md and .windsurfrules, rule priority order, per-directory scoping, and team patterns for using both files together.
Complete syntax guide for Claude Code Workflow scripts. Covers agent(), parallel(), pipeline(), phase(), schema outputs, and real patterns for multi-agent orchestration.
How to use AGENTS.md and .clinerules together in Cline. Covers native support for all rule formats, directory structure, conditional rules, and team integration patterns.
Complete guide to writing AGENTS.md — how it differs from CLAUDE.md, what Claude Code, OpenAI Codex, and GitHub Copilot each support, and copy-paste templates for Python, TypeScript, and Go projects.
Production-ready CLAUDE.md template for Python projects — virtual environment setup, ruff/mypy enforcement, pytest conventions, uv and poetry support, and PostToolUse hooks that give Claude Code instant feedback.
Production-ready CLAUDE.md templates for Vue 3 and Nuxt 3 — Composition API enforcement, Pinia conventions, auto-import boundaries, and path-scoped rules for components and composables.
Five ways claude -p fails in automation: auth key priority traps, process hangs, JSON output leaks, --bare mode bugs, and MCP limits. Exact fixes included.
Production CLAUDE.md templates for Swift/iOS — SwiftUI architecture rules, async/await conventions, XCTest patterns, SwiftLint integration, and what Claude Code can and can't do in Xcode projects.
Go向けCLAUDE.md完全テンプレート。エラーハンドリング規約・ゴルーチン安全性ルール・golangci-lint統合・コンテキスト伝播・モジュールワークフロー自動化まで網羅。
Complete CLAUDE.md template for Go — error handling conventions, goroutine safety rules, golangci-lint integration, context propagation, and module workflow automation with Claude Code.
Complete CLAUDE.md template for Rails — soft-delete awareness, RSpec conventions, Hotwire rules, path-scoped rules for models and controllers, and ActiveRecord query guardrails.
Rustプロジェクト用のプロダクション対応CLAUDE.mdテンプレートを完全解説。エラーハンドリング規則・Clippy強制・unsafeブロックポリシー・rust-analyzer連携・非同期パターンまでカバー。
Production-ready CLAUDE.md template for Rust — error handling rules, clippy enforcement, unsafe block policies, rust-analyzer integration, and async patterns that actually work with Claude Code.
Side-by-side comparison of Kiro steering files, AGENTS.md, and CLAUDE.md — syntax, scope, conditional includes, cross-tool compatibility, and a migration guide for teams moving to Kiro.
Everything you need to know about Kiro IDE — steering files, specs, hooks, and how it compares to Cursor and Claude Code. With working .kiro/steering/ examples.
Side-by-side syntax reference for all four AI coding rule formats: file location, frontmatter fields, loading scope, hierarchy support, and portability guide.
4つのAIコーディングルール形式を徹底比較——ファイルの配置、frontmatterフィールド、スコープ制御、階層サポート、ポータビリティガイドまで1ページで確認できる構文リファレンス。
FastAPI projects have specific async patterns, dependency injection conventions, and testing requirements that AI agents get wrong without explicit guidance. Here's the complete AGENTS.md playbook — copy-paste ready templates included.
.claude/agents/ YAMLフロントマターによるオーケストレーターとサブエージェントの設定を解説。ツールスコーピング、モデル選択、メモリ分離、よくある設定ミスまで網羅。
How to configure orchestrator and subagent roles using .claude/agents/ YAML frontmatter — tool scoping, model overrides, memory isolation, and common mistakes.
AiderチームはAGENTS.mdとCONVENTIONS.mdどちらを使うべきか？両者の仕組み、移行方法、マルチツール環境でAGENTS.mdが勝る理由を解説する2026年版ガイド。
Should Aider teams use AGENTS.md or CONVENTIONS.md? How both work, how to migrate, and why AGENTS.md wins for multi-tool repos in 2026.
Claude CodeをReact・Next.js開発で活用するための完全ガイド。App Router／Pages Router対応のCLAUDE.mdテンプレート、コンポーネント生成規則、TypeScript・Tailwind・Vitest・Playwrightの設定、チーム向けAGENTS.mdパターンまでコピペですぐ使える形で解説。
Complete guide to using Claude Code with React and Next.js — copy-paste CLAUDE.md templates for App Router and Pages Router, component generation patterns, TypeScript conventions, Tailwind, Vitest, Playwright, and team AGENTS.md setups.
Claude Tag brings a shared @claude teammate to Slack. This guide covers setup, access bundles, ambient mode, migration from the legacy app (deadline: August 3), and what Claude Tag cannot do.
CursorがどのようにAGENTS.mdを読み込むか、.cursor/rules（.mdc形式）との使い分け、モノレポ構成でのパターン、.cursorrulesからの移行方法を解説する2026年版完全ガイド。
How Cursor reads AGENTS.md, discovery order, combining with .cursor/rules/ .mdc files, monorepo patterns, and migration from .cursorrules — complete 2026 setup guide.
Learn exactly how .github/copilot-instructions.md works, what GitHub Copilot does with it, how it differs from AGENTS.md and CLAUDE.md, and 10 illustrative templates for teams using Copilot in 2026.
How OpenCode reads AGENTS.md, global vs project-level rules, opencode.json configuration, /init auto-generation, Build vs Plan mode — complete 2026 setup guide for the 160K-star terminal AI agent.
A production-ready CLAUDE.md template for NestJS projects, covering module structure, dependency injection rules, DTO validation patterns, TypeORM/Prisma integration, and Jest testing workflows for Claude Code.
FastAPIプロジェクト固有の非同期パターン・依存性注入・テスト規則をAIエージェントが守れるよう、AGENTS.mdに書くべき内容を完全解説。コピペ即使えるテンプレート付き。
OWASPに準拠したCLAUDE.mdルールでAIコーディングワークフローを安全にする。インジェクション防止、認証検証、シークレット管理、セキュアなデフォルト設定をカバーするコピペ可能なテンプレート。
Flutter開発でClaude Codeを最大活用するための完全なCLAUDE.mdテンプレートを提供。クリーンアーキテクチャ、Riverpodによる状態管理、iOS/Android対応のプラットフォーム固有ルールを網羅。
Claude Codeのヘッドレス-pフラグ、--output-format text/json/stream-json、--json-schemaによる構造化出力、--bareモード、CI/CD自動化のシェルスクリプトパターンを完全解説。
A no-hype breakdown of GEO tools for ChatGPT, Perplexity, Claude, and Google AI Overviews — including what separates tools that track citations from tools that help you earn them.
Generative Engine Optimization（GEO）ツールの実質的な比較。エンジン別の違い、AI Bot Analytics、クロール診断まで踏み込んで解説。
Complete guide to Claude Code's headless -p flag, --output-format text/json/stream-json, --json-schema structured output, --bare mode, and practical shell script patterns for CI/CD automation.
.cursorrulesの変換だけではない、IDEの習慣・MCP設定・Git統合・キーバインディング・メンタルモデルの転換まで含む完全なワークフロー移行ガイドです。
Claude Codeのエージェントチーム機能を使って並列開発を実現する方法を解説します。実験的機能の有効化、チーム設定、Git調整、CLAUDE.md最適化、サブエージェントとの使い分けまで実践的に解説します。
モノレポ環境でClaude Codeを効果的に使うための設定方法を解説します。ネストされたCLAUDE.mdファイルの設計、パッケージ別サブエージェントの作成、トークンバジェット管理、チームへの展開パターンを実例付きで紹介します。
Claude CodeとCursorを実際のベンチマークデータとワークフローの違いから比較します。アーキテクチャ・精度・速度・コスト・使い分けの判断基準、そして多くの開発者が採用しているハイブリッド構成まで解説します。
2026年版Claude Codeインストール手順を全プラットフォームに対応して解説します。ネイティブインストーラー・npm・Homebrew・WinGetの方法別解説、よくあるエラーの対処法、AGENTS.md/CLAUDE.mdの初期設定まで。
Claude Codeのエージェントビューを徹底解説。/bgコマンドでのセッションバックグラウンド化、claude agentsダッシュボードの操作、ワークツリー隔離の仕組み、CLIコマンド一覧、サブエージェントとの使い分けまで。2026年公式ドキュメント準拠。
Everything about Claude Code's agent view: dispatch tasks with /bg, manage parallel sessions with claude agents, understand worktree isolation, use CLI commands (attach/logs/stop), and decide when background sessions beat subagents.
Everything you need to connect Claude Code to Supabase: MCP server configuration, copy-paste CLAUDE.md templates for TypeScript/Next.js projects, safe migration patterns, RLS policy instructions, and Edge Functions development workflows.
A deep technical walkthrough of Career-Ops, the Santiago Fernández skill bundle for Claude Code. Architecture, file-by-file analysis, installation, customization, and the real workflow it enables for ML engineers.
A definitive guide to configuring Claude Code for Python projects. Copy-paste-ready CLAUDE.md and AGENTS.md templates for uv/pip, pytest, ruff, mypy, FastAPI, Django, and data science workflows.
12 production-tested best practices for Claude Code Subagents in 2026. Covers definition design, tool restrictions, model selection, context management, error handling, and cost optimization — written by practitioners running subagents in real workflows.
The wrong file choice wastes your context budget and trains bad AI habits. 11 use-case templates with file selection rationale, ready-to-copy configurations, and a composition guide for multi-tool setups.
How Claude Code Auto Mode works under the hood — the 2-stage classifier pipeline, how it differs from --dangerously-skip-permissions, customization via settings.json, and when to actually use it.
Claude Code のチェックポイントと /rewind で安全にミスを巻き戻す方法を解説。全リストア・要約オプション、Bash の制限、実用パターンまで。
How to use Claude Code checkpoints and /rewind to safely undo mistakes, explore alternatives, and compress session context. Covers all restore and summarize modes, Bash limitations, and real-world recovery patterns.
Claude Code VS Code 拡張機能の全機能解説。@メンション、IDE MCPサーバー、診断情報取得、チェックポイントrewind、/mcp管理、キーボードショートカット、ブラウザ自動化まで。2026年6月版。
Everything about the Claude Code VS Code extension — @-mentions, IDE MCP server, diagnostics, checkpoint rewind, /mcp management, keyboard shortcuts, and browser automation. Official docs reference, June 2026.
Claude Codeのセッション管理を完全解説。--continue, --resume, --fork-session, セッション命名, インタラクティブピッカー, チェックポイントまで網羅。中断したセッションをいつでも正確に再開する方法。
Complete guide to Claude Code session management in 2026. Master --continue, --resume, --fork-session, session naming, the interactive picker, and checkpointing to recover any interrupted session.
Running Claude Code in GitHub Actions and hitting permission errors? This guide covers exactly why they happen and how to fix them — with three proven patterns including allowlist config, bypassPermissions mode, and the official claude-code-action.
How to run Claude Code PR reviews in GitHub Actions using claude -p — even when reviewers can't install Claude locally. Covers diff passing, automated comment posting, review prompts for security and performance, and cost control with max-turns.
A CLAUDE.md that's too long defeats its own purpose. Here's how to use Claude Code's hierarchical CLAUDE.md system and Import directives to split a bloated root file into a structure that Claude actually reads.
Running claude -p in a pipeline and getting malformed JSON? Here's exactly why it happens, what --output-format json actually does, and how to build a Python implementation with validation and retry.
Multiple engineers using Claude Code in the same repo will eventually get a settings.json conflict. Here's exactly what to commit, what to gitignore, and how to separate team config from personal config.
Complete guide to building production AI agents with the Claude Agent SDK. Covers the query() loop, built-in tools, hooks for audit logging, subagent delegation, session continuity, MCP integration, and permission control with real code examples.
Set up Claude Code in GitHub Actions with the v1.0 GA release. Covers @claude mentions in PRs, automated code review, Bedrock and Vertex AI deployment, skills integration, security best practices, and cost optimization for CI pipelines.
Claude Code now runs natively on Windows — no WSL, no Linux subsystem. Here's the exact install process for PowerShell and CMD, how Git for Windows affects the Bash tool, and when WSL still makes sense.
Claude Code hooks stop the entire operation when they exit non-zero — here's why that happens, how each hook type behaves differently, and how to design hooks that fail gracefully instead of freezing your workflow.
Step-by-step setup for the three most useful MCP servers with Claude Code: GitHub (PR reviews, issues), Sentry (error triage), and PostgreSQL (direct database queries). Covers OAuth, headersHelper, scope management, and production configuration.
Complete guide to Claude Code Routines — cloud-scheduled automation that runs when your laptop is closed. Covers schedule triggers, API endpoints, GitHub event triggers, branch permissions, and connector configuration.
Six confirmed bugs in Claude Opus 4.8 inside Claude Code: 2-3x token regression, malformed tool-use blocks, missing responses, streaming stalls, false verification claims, and hallucinated tool outputs. Symptoms, workarounds, and whether to downgrade.
Master Claude Code's two-layer memory system: CLAUDE.md files for persistent instructions and auto memory for automatic learnings. Includes .claude/rules/, path-scoped rules, the /memory command, and InstructionsLoaded hooks.
The definitive reference for Claude Code's 6 permission modes (default, plan, acceptEdits, auto, dontAsk, bypassPermissions), rule syntax with wildcards, managed policy for teams, and hooks-based enforcement.
Orchestrator-worker, parallel reviewers, pipeline chains — learn 6 proven multi-agent patterns for Claude Code with real AGENTS.md configs and cost-efficiency benchmarks.
Stop committing .env files. This guide covers three production patterns for vaulting Claude Code secrets in 1Password — op run runtime injection, shell integration (zsh/fish/bash), and the 1Password MCP server — with a decision matrix against Doppler, Infisical, HashiCorp Vault, and SOPS.
Three config files, three different jobs. This is the definitive 2026 breakdown of scope, syntax, tool support, and which file wins for Claude Code, Cursor, and Codex CLI workflows.
Run 3–5 Claude Code sessions simultaneously without file conflicts. Covers --worktree flag, isolation: worktree for subagents, .worktreeinclude, tmux setup, and merge strategies.
Go beyond desktop notifications. This guide covers 12 production-ready hook patterns — security gates, auto-linting, Slack alerts, audit logs — with copy-paste settings.json snippets.
Plan Mode lets Claude explore your codebase before writing a line. Learn Shift+Tab, /plan, headless flags, plansDirectory, and the Explore→Plan→Implement workflow with real examples.
Skills auto-invoke based on context; slash commands need manual triggers. Learn when each wins, how to combine them, and 10 real-world patterns for your 2026 workflow. See working examples.
Master Claude Code's Plugin Marketplace introduced in 2026. Learn how to discover and install plugins, build your own plugin with skills/hooks/commands, package and publish to marketplaces, and manage scoping.
Complete configuration reference for .mcp.json and claude_desktop_config.json. Covers stdio/HTTP/SSE transports, environment variables, scope hierarchy, security settings, and 15+ real-world examples for popular MCP servers.
Complete architectural reference for Claude Code Agent Teams. Covers team lead/teammate model, shared task list, mailbox messaging, Hooks integration, cost analysis, and production patterns with real examples.
Master Claude Code Hooks with this complete reference covering all 25+ lifecycle events, exit codes, HTTP/Prompt hooks, Agent Teams integration, CI/CD patterns, and security gates. Updated for May 2026.
Set up and configure MCP servers in Claude Code for GitHub, databases, browsers, and custom tools. Covers transport types, authentication, security best practices, and troubleshooting.
Learn how to create, install, and share Claude Code skills. Covers skill anatomy, frontmatter options, tool restrictions, plugin packaging, and real-world examples for automating code review, deployment, and testing.
Set up Claude Code GitHub Actions for automated PR reviews, issue-to-code workflows, and CI/CD intelligence. Covers v1 GA configuration, Bedrock/Vertex setup, cost optimization, and real workflow examples.
Claude Code and OpenAI Codex CLI compared on architecture, cost, context window, agent systems, and real development workflows. Updated for May 2026 with Opus 4.7 and Codex GA.
Career-Ops (44,500+ stars), LazyApply, and JobScan AI are the three dominant AI job search tools in 2026. This comparison breaks down their architecture, pricing, target users, control surface, and the exact scenarios where each wins. Includes decision framework, FAQ, and migration guide between them.
Career-Ops is the viral Claude Code skill bundle (44,500+ stars in May 2026) that turned one engineer's job search from 740 listings into 1 signed Head of Applied AI offer. This is the 30-second definition, plus what makes it different, who it's for, and where to go next.
The complete, structured reference for Claude Code subagents — built-in (Explore, Plan, general-purpose), custom subagent creation, frontmatter fields, scope priority, tool restrictions, persistent memory, /agents command, hooks, model selection, forking conversations. Based on the official Anthropic May 2026 docs.
Career-Ops ships 12 distinct skill modes — from URL scanning across 45+ portals to A-F evaluation with 10 weighted dimensions, parallel batch processing via claude -p workers, and STAR interview prep accumulation. This comparison covers what each mode does, when to use it, output format, and how the modes chain together. Based on the official santifer/career-ops repository as of May 2026.
Claude Code can load AGENTS.md directly from v2.1.277, with important CLAUDE.md precedence and session exceptions. Learn when to keep @AGENTS.md and how to verify loaded instructions. Checked October 6, 2026.
Caveman is the AGENTS.md ruleset and Claude Code plugin that compresses AI agent outputs by an average of 65% without losing technical accuracy. Full breakdown of the 4 compression levels, 10 rules, benchmarks, cavecrew multi-agent pattern, /caveman-stats measurement, OpenClaw/opencode integration, and how it fits with AGENTS.md, CLAUDE.md and the broader rule-file ecosystem.
Every Claude Code plugin component in one page — verified against Anthropic's May 2026 docs. Cover Skills, Agents, Hooks, MCP servers, LSP servers, Monitors, Commands, the plugin.json schema, .claude-plugin/ structure, marketplace distribution, and production patterns.
Step-by-step tutorial for using Career-Ops — the viral Claude Code job-search system. CV onboarding, portal scanning, batch evaluation, application tracking, and the exact prompts to drive each of the 14 skill modes. Verified against the GitHub source as of May 2026.
Windsurf Rules（global_rules.md）、Cursor Rules（.cursorrules/MDC形式）、AGENTS.mdを横断比較。スコープ・構文・マイグレーション手順を解説。
Every Claude Code hook event in one page — verified against Anthropic's May 2026 docs. Cover all 32+ event types, 5 handler types (command/http/mcp_tool/prompt/agent), matcher patterns, exit code 2 blocking semantics, JSON input/output schemas, and production patterns.
Roo Codeのカスタムインストラクション完全解説。.roo/rules/ディレクトリの設定、カスタムモード（Architect、Debug、Code）の作成方法、モード別ルールの書き方を実例付きで解説します。
Complete guide to Roo Code custom instructions. Learn how to set up .roo/rules/ directory, create custom modes (Architect, Debug, Code), and write mode-specific rules for your project.
AGENTS.md, CLAUDE.md, and .cursor/rules each solve AI coding configuration differently. We compare all three on scope, frontmatter, scoping, imports, and rule activation — verified against Anthropic and Cursor 2026 docs — plus migration paths between them.
Every Claude Code subagent setting in one place — verified against Anthropic's May 2026 docs. Cover 17 frontmatter fields, 5 scope locations, 3 built-in subagents (Explore/Plan/general-purpose), permission modes, MCP scoping, and isolation:worktree.
Santiago's Career-Ops hit 44,554+ GitHub stars (44.5K within a week). Beyond the job-search use case, it's a reference implementation for building production Claude Code Skills. We dissect the 14-mode architecture, AGENTS.md setup, 10-dimensional scoring, and what you can steal for your own systems.
Career-Ops by Santiago turned 740 listings into 1 Head of Applied AI offer. Full reference to 14 skill modes, the A–G evaluation rubric, multi-agent design, and how 3 community forks differ — all verified against the GitHub source.
Step-by-step Claude Code installation guide for 2026. Native installer, npm, Homebrew, WinGet — covering all platforms, common errors, AGENTS.md/CLAUDE.md setup, and first-run configuration.
2026 Q2 updated comparison of GitHub Copilot CLI and Claude Code CLI. Model availability (Opus 4.7 / GPT-5 / Gemini 3 Pro), agent architecture, pricing tiers, SWE-bench scores, AGENTS.md compatibility, ideal use cases, and how senior engineers compose both. FAQ included for AGENTS.md and pricing questions.
Complete 2026 guide to building production-grade agentic AI systems. Coordinator-specialist patterns, AGENTS.md/CLAUDE.md integration, governance, observability, and human-in-the-loop design from Anthropic, Google, and Shopify.
Deep analysis of AGENTS.md files from n8n, awesome-go, LangFlow, Deno, Linear, Grafana, OpenAPI Generator, and Astro. What they have in common, what makes each unique, and what every developer can learn from them.
Production-ready CLAUDE.md templates for Python projects. Covers modern Python toolchain (uv, ruff, pyproject.toml, pytest), with separate patterns for FastAPI, Django, and data science projects. Copy-paste ready.
Design principles and operational patterns for AGENTS.md drawn from 60,000+ real repositories. Beyond format basics — how top projects structure instructions, handle multi-tool environments, and keep AGENTS.md from going stale.
A technical deep-dive comparing CLAUDE.md, CONVENTIONS.md, and AGENTS.md — format, supported tools, strengths, weaknesses, migration paths, and real code examples. Decide which AI coding rules format belongs in your repo.
Concrete CLAUDE.md best practices drawn from analyzing 100+ real-world repositories. 12 actionable patterns, common anti-patterns, and why most CLAUDE.md files underperform. For developers who want results, not theory.
A rigorous methodology-first benchmark of the four leading AI coding tools — Cursor, Claude Code, Aider, and OpenAI Codex CLI — tested across five real-world task categories. With cost-per-task tables, hidden overhead analysis, and a recommendation matrix by use case.
A complete comparison of OrbStack and Docker Desktop for Mac developers in 2026. Real benchmarks on M2 MacBook Pro, pricing at team scale, Kubernetes support, Apple Containers context, and a clear decision guide for every use case.
A decision-focused deep dive into when to use sub-agents versus Agent Teams, with a full pattern catalog, cost and token analysis, failure mode taxonomy, and the honest case for not going multi-agent at all.
AGENTS.md を書いたら効いていると思い込んでいないか？指示の遵守率を測定し、欠如時との差分を計測し、サイレントに失敗しているルールを修正する系統的アプローチを解説する。
Most teams write AGENTS.md and assume it works. Here's a systematic approach to testing whether your instructions are followed, measuring the delta when they're absent, and fixing the rules that silently fail.
Going beyond cost optimization — learn how to actually see what Claude Code is doing. Hooks-based telemetry, JSONL transcript parsing, LangSmith/Langfuse integration, Grafana dashboards, and production-grade error tracing for Claude Code in 2026.
Claude Code がサブエージェントを生成すると、全員が同じ AGENTS.md を読む。でも専門特化したエージェントには、その役割に必要なコンテキストだけを渡すべきだ。エージェントごとの AGENTS.md 設計パターンを解説する。
When Claude Code spawns sub-agents, they all read the same root AGENTS.md — but that's often the wrong instructions for a specialized agent. Here's how to design per-agent AGENTS.md files that give each sub-agent the context it actually needs.
AGENTS.md はバージョン管理に入り、チームで共有され、公開リポジトリに上がることもある。何を書くべきで何を書くべきでないか、プロンプトインジェクション攻撃の実態まで具体的に解説する。
AGENTS.md files end up in version control, shared across teams, and sometimes public repositories. Here's a concrete guide to what belongs in them, what doesn't, and how prompt injection via AGENTS.md actually works.
Gemini CLIがAGENTS.mdをサポートした。でも「互換性あり」は表面的な話。Claude Codeとどう解釈が違うのか、2ツール対応の書き方はどうすべきかを具体的に解説する。
Gemini CLI landed with its own take on AGENTS.md support. Here's exactly what it reads, what it ignores, and how to write instructions that work across both Gemini CLI and Claude Code without maintaining two files.
A practical guide to using Claude Code, Cursor, Codex CLI, and Aider for mobile development across Flutter, React Native, Swift/SwiftUI, and Kotlin/Compose — with CLAUDE.md and AGENTS.md templates for each framework.
AGENTS.mdにはスキーマも強制フォーマットもツールチェーンもない。ASTスタイルのパース・ルール定義・Pre-commitとCIに組み込めるCLIまで、軽量バリデーターをゼロから構築する。
AGENTS.md files have no schema, no enforced format, and no tooling. Here's how to build a lightweight validator from scratch — with AST-style parsing, rule definitions, and a CLI you can wire into pre-commit and CI.
Plan modeはコードレビューだけのものではない。CLAUDE.mdを1行も書く前に設計するための最適なツールだ。矛盾・ギャップ・過剰仕様を問題になる前に洗い出す実践的なワークフロー。
Plan mode isn't just for code review — it's the right tool for designing CLAUDE.md before you write a single line. Here's a practical workflow for using Claude's planning pass to surface conflicts, gaps, and over-specification before they become problems.
AGENTS.mdは静かに陳腐化する。GitHub ActionsとPre-commitフックでドリフトを検知し、本番環境に影響が出る前にキャッチする実践的なワークフローYAMLとシェルスクリプト。
AGENTS.md is a living document that drifts silently. Here's how to wire GitHub Actions, pre-commit hooks, and automated review bots to catch staleness before it hits production — with real workflow YAML and shell scripts.
Monorepos break single-file AI instruction conventions. Here's how to use hierarchical AGENTS.md files across packages, workspaces, and CI contexts — with real Nx, Turborepo, and pnpm workspace patterns.
モノレポ環境では単一のAGENTS.mdでは対応しきれない。Nx・Turborepo・pnpmワークスペースを使った階層的AGENTS.md運用と、重複管理のコストを下げる実践パターンを解説。
CursorのProject Rules（.mdc形式）とAGENTS.mdは似た問題を異なるトレードオフで解決する。フォーマット、スコーピング、アクティベーション、デュアルスタック構成まで実例で比較。
Cursor's .mdc Project Rules and AGENTS.md solve similar problems with different trade-offs. Here's a direct comparison of file formats, scoping behavior, activation logic, and the dual-stack setup that gets you the best of both.
CLAUDE.mdのトークン消費を40%以上削減する実践テクニック。@import構文、ファイル分割パターン、プロンプトキャッシュとの相性、実測削減量の計測方法まで解説。
Cut your CLAUDE.md token consumption by 40%+ without losing functionality. Covers @import syntax, file splitting patterns, lazy loading, prompt cache compatibility, and how to measure what you actually cut.
Stop choosing between Cursor and Claude Code. The developers shipping the most in 2026 are running both — in a deliberate hybrid workflow that plays to each tool's strengths. Here's the complete playbook.
You have Codex CLI installed and AGENTS.md configured. Now what? This guide catalogs 7 proven workflow patterns — inline review, spec-driven generation, test-first refactor, multi-step planning, and more — plus AGENTS.md tuning, a feature comparison against Claude Code and Cursor, cost optimization, and the pitfalls that catch most developers off guard.
GitHub CopilotはAGENTS.mdをどう読むのか（2026年Q2更新版）。Copilot Chat / Workspace / Coding Agent / CLI の挙動差、.github/copilot-instructions.md とAGENTS.md の使い分け、Claude Codeとの共存設定、トークン予算測定、FAQまで実例で解説。
Does GitHub Copilot actually read AGENTS.md? 2026 Q2 updated guide covering Copilot Chat / Workspace / Coding Agent / CLI behavior, .github/copilot-instructions.md vs AGENTS.md tradeoffs, and configuration patterns for mixed Copilot + Claude Code teams. FAQ + production token-budget measurements included.
OpenAI Codex エージェント向けの AGENTS.md 執筆技術ガイド。Codex が実際に読むもの・CLAUDE.md との動作の違い・CI パイプライン連携・一貫した結果を生むパターンを解説。
A technical guide to writing AGENTS.md for OpenAI Codex agents. Covers what Codex actually reads, differences from CLAUDE.md behavior, CI pipeline integration, and patterns that produce consistent results.
A deep dive into when to use AGENTS.md vs CLAUDE.md — with a full decision matrix, real repo walkthroughs, migration patterns, coexistence strategies, and 7 pitfalls to avoid. For devs who have already read the basics and want the full picture.
Your .cursorrules file works, but it's locked to Cursor and ignored in Agent mode. Here's the complete migration path to AGENTS.md — conversion script, gotchas, and how to keep both working during the transition.
.cursorrulesはCursor ChatとComposerでは動くが、Agent modeでは無視される。AGENTS.mdへの移行手順、変換スクリプト、Cursor互換性の維持方法、よくあるハマりどころを解説。
一貫した AI 支援開発のための Dev Container と Claude Code の組み合わせ設定。devcontainer.json・コンテナ環境向け CLAUDE.md・コンテナ内 MCP サーバー・チームオンボーディングまで解説。
How to configure Dev Containers alongside Claude Code for consistent AI-assisted development. Covers devcontainer.json setup, CLAUDE.md for container environments, MCP servers in containers, and team onboarding.
When your monorepo grows past 50 packages, standard AI tool setups break down. This deep-dive covers 7 configuration patterns, per-tool walkthroughs for Nx, Turborepo, pnpm workspaces, and Bazel, plus token budget measurement, CI/CD integration, and the pitfalls that only appear at scale.
Claude Code をプロジェクト固有のツールで拡張するカスタム MCP サーバーの構築方法。TypeScript SDK・ツールスキーマ設計・リソースエンドポイント・エラーハンドリング・settings.json による接続まで解説。
How to build a custom MCP server that extends Claude Code with project-specific tools. Covers the TypeScript SDK, tool schema design, resource endpoints, error handling, and connecting to Claude Code via settings.json.
Anthropic bills by the token, and Claude Code can burn through them fast. This guide covers model selection, prompt caching, CLAUDE.md design, .claudeignore, worktrees, and sub-agent patterns that keep costs under control without slowing you down.
Claude Code で体系的なコードレビューを実施する方法。PR分析・セキュリティチェック・アーキテクチャフィードバック、そしてレビューモードと実装モードで動作を変えるCLAUDE.md設定まで解説。
How to use Claude Code for systematic code review — PR analysis, security checks, architectural feedback, and how to configure CLAUDE.md so review mode behaves differently from implementation mode.
TypeScript プロジェクト向けに最適化した CLAUDE.md テンプレート。型安全性・インポート規約・エラーハンドリング・テスト基準・禁止パターンを網羅したプロダクション対応版。
A complete CLAUDE.md template tuned for TypeScript projects. Covers type safety rules, import conventions, error handling patterns, testing standards, and common anti-patterns to block.
Claude Code のパーミッションシステムを徹底解説。信頼レベル・許可ツール・危険操作・settings.json によるプロジェクト別の許可/拒否リスト設定まで網羅。
Understand how Claude Code's permission system works: trust levels, allowed tools, dangerous operations, and how to configure project-specific allow/deny lists in settings.json.
.clinerules と Roo Code の .roo/rules/ を技術的に比較。ファイル構造・グロブマッチング・モード別設定・MCPパーミッションの違いから、プロジェクトに合った選択肢を解説します。
A deep comparison of .clinerules and Roo Code's .roo/rules/ system. Covers file format, glob matching, mode-specific config, context budget, and which tool fits your workflow in 2026.
Set up Claude Code for Flutter development with a complete CLAUDE.md template. Covers clean architecture, Riverpod state management, and platform-specific rules for iOS/Android.
Secure your AI coding workflow with OWASP-aligned CLAUDE.md rules. Copy-paste template covering injection prevention, auth validation, secrets management, and secure defaults.
Side-by-side comparison of Windsurf Rules (global_rules.md), Cursor Rules (.cursorrules/MDC), and AGENTS.md. Learn the scope, syntax, and migration path for each AI coding config format.
PythonプロジェクトのCLAUDE.md・.cursorrulesテンプレートを徹底解説。型ヒント、テスト規約、パッケージ管理、非同期パターンのルール設定でAIアシスタントの出力品質を向上。
Complete guide to writing AI coding rules for Python projects — CLAUDE.md templates, .cursorrules examples, type hints, testing conventions, packaging, and environment management for Claude Code and Cursor.
React・Next.jsプロジェクト向けの実用的なAIコーディングルール。サーバー/クライアントコンポーネントの判断基準、App Router規約、TypeScriptパターン、テスト設定のテンプレートを提供。
Complete AI coding rules for React and Next.js projects — server vs client component rules, App Router conventions, TypeScript patterns, testing with Vitest, and production-ready CLAUDE.md templates.
Claude Codeの記憶システムを実際に機能させる設計と運用方法。MEMORY.mdアーキテクチャ、セッション引き継ぎワークフロー、Auto Memoryの管理、チーム・個人開発者向けパターンを解説。
How to design and maintain a working Claude Code memory system — MEMORY.md architecture, session handoff workflows, Auto Memory curation, and patterns for teams vs. solo developers.
Claude Code サブエージェントのベストプラクティスを2026年版で網羅。実本番のAGENTS.md 12パターンを分析。タスク分解、コンテキスト受け渡し、並列実行、エラー復旧、見た目は正しいが大規模化で崩れる失敗モードまで。
Claude Code subagents best practices in 2026 — 12 production-tested patterns from real AGENTS.md files. Task decomposition, context passing, parallel execution, error recovery, and the failure modes that look fine until they don't.
基本的な削減を超えたCLAUDE.mdの高度なトークン最適化技術。レイジーロード、条件付きセクション、メモリオフロード、実測方法を解説します。
Advanced token optimization strategies for CLAUDE.md beyond basic pruning — lazy loading, conditional sections, memory offloading, and real measurement techniques to reclaim your Claude Code context window.
CLAUDE.mdを効果的に書くためのステップバイステップガイド。何を書くべきかの特定方法、Claudeが実際に従う命令の書き方、CLAUDE.mdを悪化させる典型的なミスとその修正方法を解説。
A step-by-step process for writing effective CLAUDE.md files — how to identify what to include, how to phrase instructions Claude actually follows, and common mistakes that make CLAUDE.md files worse instead of better.
OpenAI Codex CLIでAGENTS.mdを活用する方法を解説（2026年Q2更新版）。ファイル配置、構文、32KiB制限、override機構、モノレポパターン、実プロジェクトの設定例、ベストプラクティスまで網羅。Codex CLI最新挙動・spec安定化対応。
Complete 2026 Q2 guide to writing and configuring AGENTS.md for OpenAI Codex CLI. Covers file discovery, the 32 KiB limit, override mechanism, monorepo patterns, real production examples, and best practices for agentic coding. Updated for spec stabilization and new Codex CLI behavior.
.cursorrulesと.cursor/rules（MDC形式）の違いを徹底比較。構文の違い、スコープ管理、移行手順、プロジェクト規模別の選び方まで解説。
Compare .cursorrules and .cursor/rules MDC format side by side. Covers syntax differences, scoping, migration steps, and when to use each format in your Cursor AI workflow.
Everything you need to know about .windsurfrules — how they work, how they differ from .cursorrules and CLAUDE.md, and 12 practical examples you can use today.
AGENTS.md is becoming the standard for multi-agent AI instruction files. Learn exactly how to write one that actually works — with real examples, anti-patterns to avoid, and tested patterns for orchestrators, subagents, and hybrid setups.
How to wire Claude Code's hook system into pre-commit so linting, tests, and secret scanning run automatically before every commit. Includes copy-paste configs for Python, TypeScript, and polyglot repos.
A practical debugging workflow for Claude Code. Covers error reproduction, Plan Mode for complex bugs, test-driven debugging, hooks for auto-linting, and the /rewind escape hatch.
Learn how to connect Claude Code to external tools via MCP servers. Covers configuration, scoping, security, the top servers worth installing, and how Tool Search keeps context usage manageable.
A practical comparison of Claude Code and Cursor based on real benchmark data and workflow differences. Covers architecture, speed, accuracy, cost, and when to use each tool — including the hybrid setup most experienced developers have landed on.
Learn how to configure Cline with .clinerules files. Covers single-file and directory formats, YAML frontmatter, glob-based conditional rules, and 5 copy-pasteable templates.
Set up Claude Code Agent Teams for real parallel development. Covers the experimental feature, team configs, git coordination, CLAUDE.md optimization, and when to use it.
Copy-pasteable Cursor rules templates for React/Next.js, Python, Go, TypeScript, and more. Covers .cursorrules and the new .cursor/rules/*.mdc format.
A practical breakdown of Claude Code's four-layer memory system — CLAUDE.md, auto memory, MEMORY.md, and Auto-Dream — with honest coverage of what persists, what doesn't, and the patterns that actually survive session restarts.
Claude Code defaults to writing implementation first. Here is how to flip that with PreToolUse hooks, CLAUDE.md constraints, and a skill-based red-green-refactor loop that actually holds up across a full session.
How to measure, cut, and restructure your CLAUDE.md for maximum context efficiency. Covers token counting, what to prune, lazy-loading patterns, and how to reclaim 30–50% of your context window without sacrificing behavior.
Not just a .cursorrules conversion — a full workflow migration guide covering IDE habits, MCP setup, git integration, keybindings, and the mental model shift that actually makes Claude Code click.
Learn how to use git worktrees with Claude Code to run multiple AI sessions in parallel on separate branches. Covers the --worktree flag, subagent isolation, cleanup behavior, and real-world parallel task patterns.
A practical step-by-step guide to converting your existing Cursor rules to CLAUDE.md for Claude Code. Covers rule translation, what carries over, what needs rewriting, and how to support both tools from a single AGENTS.md.
A practical guide to writing effective AGENTS.md files for Claude Code and multi-agent systems. Covers format, scope hierarchy, entry design, tool permissions, and the patterns that make subagents actually useful.
Why Claude Code drains your context window faster than the 1M token limit suggests — and 7 fixes that actually work in 2026. Covers .claudeignore patterns, lean CLAUDE.md design, compaction, subagents, and session resume.
Learn how to build custom slash commands in Claude Code using the Skills system. Covers SKILL.md structure, command vs skill patterns, scope, progressive disclosure, and practical examples you can copy today.
Learn how to use Claude Code subagents effectively — when to delegate, how to write AGENTS.md entries, parallel vs sequential patterns, and the key mistake that wastes your context window.
Learn how to use Claude Code hooks to automate formatting, block dangerous commands, inject context, and enforce team rules. Includes 10 copy-paste-ready examples covering all major hook events.
Learn how to configure Claude Code for monorepos using nested CLAUDE.md files, package-specific subagents, token budget strategies, and team-friendly patterns. Includes a real-world example with a full-stack TypeScript monorepo.
Curated list of the best Claude Code Skills (custom slash commands) on GitHub, updated for 2026 Q2. From design audits to security research, session management, and subagent integration — these skills turn Claude Code into a specialized expert. Includes installation guide, FAQ, and ecosystem updates.
A complete guide to Claude Code best practices, combining Anthropic's official documentation with battle-tested patterns from the developer community. Covers CLAUDE.md, hooks, skills, memory, and workflow optimization.
AGENTS.md is the multi-tool standard (60K+ repos). CLAUDE.md is Claude Code's native config. Learn when to use each, how to combine them with @import, and avoid the 7 most common setup mistakes.
Next.js、Python ML、API、モバイル、DevOps、OSSなど用途別のCLAUDE.mdテンプレートを10個紹介。コピペしてすぐにClaude Codeを最適化できる実践的な設定集。
Ready-to-use CLAUDE.md templates for Next.js, Python ML, REST APIs, React Native, DevOps, open source libraries, monorepos, documentation sites, Chrome extensions, and CLI tools. Copy, paste, customize, ship.
A curated list of the 20 best AI coding rules from real open-source projects. Featuring CLAUDE.md, .cursorrules, and AGENTS.md files from Vercel, LangChain, Deno, Grafana, and more — with analysis of what makes each one effective.
Learn how to write an effective CLAUDE.md file for Claude Code. This 2026 Q2 update covers token-cost economics, the Anthropic Official Include/Exclude table, the @import pattern, CLAUDE.md vs Skills decision rules, real-world anti-patterns, and how to treat CLAUDE.md as code. Includes FAQ + FAQPage schema for AI referral.
Cursor Rules vs CLAUDE.md vs AGENTS.md — a complete 2026 comparison with real-world data. Which format works with which tool, how to migrate between them, and the hybrid setup most teams actually ship today.
Browse our curated collection of CLAUDE.md, .cursorrules, AGENTS.md, and image prompts.