[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-79887":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":21,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":24,"createdAt":10,"pushedAt":10,"updatedAt":45,"readmeContent":46,"aiSummary":47,"trendingCount":15,"starSnapshotCount":15,"syncStatus":48,"lastSyncTime":49,"discoverSource":50},79887,"sample-well-architected-skills-and-steering","aws-samples\u002Fsample-well-architected-skills-and-steering","aws-samples","Reusable skills and steering that teach AI coding agents how to apply the AWS Well-Architected Framework. One set of playbooks, 13 supported tools.","https:\u002F\u002Faws.amazon.com\u002Farchitecture\u002Fwell-architected\u002F",null,"Python",175,29,4,0,3,14,60,10,4.43,"MIT No Attribution",false,"main",[25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44],"ai-coding-agents","amp","antigravity","aws","aws-well-architected","claude-code","cline","codex","cursor","devops-agent","gemini-cli","github-copilot","junie","kiro","kiro-cli","kiro-ide","mcp","skills","steering","windsurf","2026-06-12 02:03:55","# 🏗️ Well-Architected Skills & Steering for AI Coding Agents\n\nReusable skills and steering that teach AI coding agents how to apply the [AWS Well-Architected Framework](https:\u002F\u002Fdocs.aws.amazon.com\u002Fwellarchitected\u002Flatest\u002Fframework\u002Fwelcome.html). One set of playbooks, **12 supported tools**.\n\n\u003Cdiv align=\"center\">\n\n**Kiro** · **Claude Code** · **Cursor** · **Codex** · **Windsurf** · **GitHub Copilot** · **Gemini CLI** · **Antigravity** · **Junie** · **Amp** · **Cline** · **AWS DevOps Agent**\n\n\u003C\u002Fdiv>\n\n> [!IMPORTANT]\n> This sample is provided for educational and demonstrative purposes. It is not intended for production use without additional review and testing appropriate to your environment.\n\n---\n\n## 🎯 Why this exists\n\nDevelopers don't stop to consult documentation — they ask their AI assistant. If the assistant doesn't know the Well-Architected Framework, the guidance never reaches the code.\n\nThis project embeds WA best practices **where development actually happens**: in the IDE, at the moment code is being written. Instead of treating architecture reviews as a separate gate, teams get continuous, contextual guidance that:\n\n- ✅ Reduces rework by catching misalignments early\n- ✅ Works across 12 AI coding tools with a single source of truth\n- ✅ Requires no AWS credentials, no API calls — everything runs locally\n- ✅ Follows the open [Agent Skills specification](https:\u002F\u002Fagentskills.io\u002F)\n\n---\n\n## 📦 What's inside\n\n```text\nsteering\u002F                           Always-on context (Kiro)\n  well-architected.md                 Pillars, design principles, review process\n\nskills\u002F                             Step-by-step playbooks (tool-agnostic)\n  wa-review\u002F                          Full review across all 6 pillars\n  security-assessment\u002F                IAM, detection, data protection, incident response\n  reliability-improvement-plan\u002F       SPOFs, recovery, scaling, change management\n  cost-optimization-audit\u002F            Waste, right-sizing, pricing models\n  performance-efficiency\u002F             Resource selection, scaling, caching\n  sustainability-optimization\u002F        Utilization, managed services, data lifecycle\n  operational-excellence\u002F             CI\u002FCD, observability, incidents, automation\n  migration-readiness\u002F                7 Rs assessment with migration plan\n  architecture-decision-record\u002F       WA-aligned ADRs with pillar impact\n\nassets\u002F                             Shared reference material\n  v13\u002F                                307 WA Framework best practices (by ID)\n  well-architected-best-practices.md  Per-pillar investigation checklists\n  cloudwatch-metrics-reference.md     Metric thresholds + composite alarm patterns\n  incident-investigation-patterns.md  Triage, RCA, mitigation playbooks\n  skill-authoring-guide.md            DevOps Agent skill authoring guide\n\nadapters\u002F                           Tool-specific configuration\n  claude-code\u002F                        CLAUDE.md + slash commands\n  cursor\u002F                             .cursor\u002Frules\u002F*.md\n  codex\u002F                              AGENTS.md\n  windsurf\u002F                           .windsurfrules\n  github-copilot\u002F                     .github\u002Fcopilot-instructions.md\n  cline\u002F                              .clinerules\n  gemini-cli\u002F                         GEMINI.md\n  antigravity\u002F                        .agents\u002Frules\u002F*.md\n  junie\u002F                              .junie\u002Fguidelines + .junie\u002Fskills\n  amp\u002F                                .agents\u002Fskills\u002F*.md\n  devops-agent\u002F                       Packaging for AWS DevOps Agent\n\npowers\u002F                             Kiro Powers (coming soon)\n\nevals\u002F                              Automated evaluation runner (Bedrock)\n  run.py                              CLI entry point\n  grade.py                            LLM-as-judge grader\n  report.py                           Scoring and terminal output\n  config.yaml                         Bedrock region and model config\n  pyproject.toml                      Dependencies (use uv sync)\n\ninstall.sh                          One-command setup (macOS\u002FLinux)\ninstall.ps1                         One-command setup (Windows PowerShell)\n```\n\n---\n\n## 🚀 Quick start\n\n### One-liner (no clone needed)\n\n#### Via [skills.sh](https:\u002F\u002Fskills.sh)\n\n```bash\nnpx skills add aws-samples\u002Fsample-well-architected-skills-and-steering\n```\n\nAuto-detects your AI agent and installs skills directly. Use `--list` to preview available skills, or `--skill \u003Cname>` to install a specific one:\n\n```bash\n# List available skills\nnpx skills add aws-samples\u002Fsample-well-architected-skills-and-steering --list\n\n# Install a specific skill\nnpx skills add aws-samples\u002Fsample-well-architected-skills-and-steering --skill wa-review\n\n# Install globally (user-level, applies to all projects)\nnpx skills add aws-samples\u002Fsample-well-architected-skills-and-steering -g\n```\n\n#### Via bootstrap script\n\n**macOS \u002F Linux:**\n\n```bash\ncurl -sL https:\u002F\u002Fraw.githubusercontent.com\u002Faws-samples\u002Fsample-well-architected-skills-and-steering\u002Fmain\u002Fbootstrap.sh | bash\n```\n\n**Windows (PowerShell):**\n\n```powershell\n& ([scriptblock]::Create((irm https:\u002F\u002Fraw.githubusercontent.com\u002Faws-samples\u002Fsample-well-architected-skills-and-steering\u002Fmain\u002Fbootstrap.ps1)))\n```\n\nAuto-detects your AI tools (`.cursor\u002F`, `.claude\u002F`, `.kiro\u002F`, `.junie\u002F`, etc.), installs for all of them, and cleans up.\n\nTo install for a specific tool instead:\n\n```bash\n# macOS \u002F Linux\ncurl -sL ...\u002Fbootstrap.sh | bash -s -- --tool kiro\n\n# Windows (PowerShell)\n& ([scriptblock]::Create((irm ...\u002Fbootstrap.ps1))) -Tool kiro\n```\n\n### Install script (from local clone)\n\n**macOS \u002F Linux:**\n\n```bash\n# Auto-detect tools in your project\n.\u002Finstall.sh ~\u002Fmy-project --tool auto\n\n# Install for a specific tool\n.\u002Finstall.sh ~\u002Fmy-project --tool claude-code\n\n# Install for multiple tools at once\n.\u002Finstall.sh ~\u002Fmy-project --tool kiro --tool claude-code --tool cursor\n\n# Install for all supported tools\n.\u002Finstall.sh ~\u002Fmy-project --tool all\n\n# Use symlinks for automatic updates\n.\u002Finstall.sh ~\u002Fmy-project --tool claude-code --symlink\n\n# Install globally (applies to all projects)\n.\u002Finstall.sh --global --tool claude-code\n```\n\n**Windows (PowerShell):**\n\n```powershell\n# Auto-detect tools in your project\n.\\install.ps1 -TargetDir C:\\Projects\\my-app -Tool auto\n\n# Install for a specific tool\n.\\install.ps1 -TargetDir C:\\Projects\\my-app -Tool claude-code\n\n# Install for multiple tools at once\n.\\install.ps1 -Tool kiro, claude-code, cursor\n\n# Install for all supported tools\n.\\install.ps1 -Tool all -Force\n\n# Install globally (applies to all projects)\n.\\install.ps1 -Global -Tool claude-code\n```\n\n> [!TIP]\n> Use `--symlink` (bash) or `-Symlink` (PowerShell) to create symbolic links instead of copies. When this repo updates, your project gets the changes automatically without reinstalling. On Windows, symlinks require elevated permissions.\n\n> [!NOTE]\n> **Global installs** place files in your home directory (`~\u002FCLAUDE.md`, `~\u002F.kiro\u002F`, `~\u002F.cursor\u002F`, etc.) and apply to all projects without their own config. Use project-level installation (the default) if you only want WA guidance for specific projects.\n>\n> **Existing files** — the installer prompts before overwriting. Use `--force` to skip confirmation.\n\n---\n\n### Manual installation\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Kiro\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\nmkdir -p .kiro\u002Fsteering .kiro\u002Fskills\ncp path\u002Fto\u002Fthis-repo\u002Fsteering\u002Fwell-architected.md .kiro\u002Fsteering\u002F\ncp -r path\u002Fto\u002Fthis-repo\u002Fskills\u002F* .kiro\u002Fskills\u002F\n```\n\nWindows (PowerShell):\n\n```powershell\nNew-Item -ItemType Directory -Force -Path .kiro\\steering, .kiro\\skills\nCopy-Item path\\to\\this-repo\\steering\\well-architected.md .kiro\\steering\\\nCopy-Item -Recurse path\\to\\this-repo\\skills\\* .kiro\\skills\\\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Claude Code\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fclaude-code\u002FCLAUDE.md .\u002FCLAUDE.md\ncp -r path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fclaude-code\u002Fcommands .claude\u002Fcommands\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item path\\to\\this-repo\\adapters\\claude-code\\CLAUDE.md .\\CLAUDE.md\nCopy-Item -Recurse path\\to\\this-repo\\adapters\\claude-code\\commands .claude\\commands\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Cursor\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp -r path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fcursor\u002Frules .cursor\u002Frules\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item -Recurse path\\to\\this-repo\\adapters\\cursor\\rules .cursor\\rules\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Codex (OpenAI)\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fcodex\u002FAGENTS.md .\u002FAGENTS.md\ncp -r path\u002Fto\u002Fthis-repo\u002Fskills .\u002Fskills\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item path\\to\\this-repo\\adapters\\codex\\AGENTS.md .\\AGENTS.md\nCopy-Item -Recurse path\\to\\this-repo\\skills .\\skills\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Windsurf\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fwindsurf\u002F.windsurfrules .\u002F.windsurfrules\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item path\\to\\this-repo\\adapters\\windsurf\\.windsurfrules .\\.windsurfrules\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 GitHub Copilot\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\nmkdir -p .github\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fgithub-copilot\u002F.github\u002Fcopilot-instructions.md .github\u002F\n```\n\nWindows (PowerShell):\n\n```powershell\nNew-Item -ItemType Directory -Force -Path .github\nCopy-Item path\\to\\this-repo\\adapters\\github-copilot\\.github\\copilot-instructions.md .github\\\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Gemini CLI\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fgemini-cli\u002FGEMINI.md .\u002FGEMINI.md\ncp -r path\u002Fto\u002Fthis-repo\u002Fskills .\u002Fskills\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item path\\to\\this-repo\\adapters\\gemini-cli\\GEMINI.md .\\GEMINI.md\nCopy-Item -Recurse path\\to\\this-repo\\skills .\\skills\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Antigravity\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\nmkdir -p .agents\u002Frules .agents\u002Fskills\ncp -r path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fantigravity\u002Frules\u002F* .agents\u002Frules\u002F\nfor skill_dir in path\u002Fto\u002Fthis-repo\u002Fskills\u002F*\u002F; do\n  skill_name=$(basename \"$skill_dir\")\n  mkdir -p \".agents\u002Fskills\u002F$skill_name\"\n  cp \"$skill_dir\u002FSKILL.md\" \".agents\u002Fskills\u002F$skill_name\u002FSKILL.md\"\ndone\n```\n\nWindows (PowerShell):\n\n```powershell\nNew-Item -ItemType Directory -Force -Path .agents\\rules, .agents\\skills\nCopy-Item -Recurse path\\to\\this-repo\\adapters\\antigravity\\rules\\* .agents\\rules\\\nGet-ChildItem path\\to\\this-repo\\skills -Directory | ForEach-Object {\n    New-Item -ItemType Directory -Force -Path \".agents\\skills\\$($_.Name)\"\n    Copy-Item \"$($_.FullName)\\SKILL.md\" \".agents\\skills\\$($_.Name)\\SKILL.md\"\n}\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Junie (JetBrains)\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\nmkdir -p .junie\u002Fguidelines .junie\u002Fskills\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fjunie\u002Fguidelines.md .junie\u002Fguidelines\u002Fwell-architected.md\ncp -r path\u002Fto\u002Fthis-repo\u002Fskills\u002F* .junie\u002Fskills\u002F\n```\n\nWindows (PowerShell):\n\n```powershell\nNew-Item -ItemType Directory -Force -Path .junie\\guidelines, .junie\\skills\nCopy-Item path\\to\\this-repo\\adapters\\junie\\guidelines.md .junie\\guidelines\\well-architected.md\nCopy-Item -Recurse path\\to\\this-repo\\skills\\* .junie\\skills\\\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Amp\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Famp\u002FAGENTS.md .\u002FAGENTS.md\nmkdir -p .agents\u002Fskills\ncp -r path\u002Fto\u002Fthis-repo\u002Fskills\u002F* .agents\u002Fskills\u002F\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item path\\to\\this-repo\\adapters\\amp\\AGENTS.md .\\AGENTS.md\nNew-Item -ItemType Directory -Force -Path .agents\\skills\nCopy-Item -Recurse path\\to\\this-repo\\skills\\* .agents\\skills\\\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 Cline\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\ncp path\u002Fto\u002Fthis-repo\u002Fadapters\u002Fcline\u002F.clinerules .\u002F.clinerules\n```\n\nWindows (PowerShell):\n\n```powershell\nCopy-Item path\\to\\this-repo\\adapters\\cline\\.clinerules .\\.clinerules\n```\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>\u003Cstrong>🔹 AWS DevOps Agent\u003C\u002Fstrong>\u003C\u002Fsummary>\n\nmacOS \u002F Linux:\n\n```bash\n# Package all skills as zip files for upload to your Agent Space\n.\u002Finstall.sh ~\u002Foutput-dir --tool devops-agent\n# Then upload each .zip from ~\u002Foutput-dir\u002Fdevops-agent-skills\u002F via the Operator Web App\n```\n\nWindows (PowerShell):\n\n```powershell\n# Package all skills as zip files for upload to your Agent Space\n.\\install.ps1 -TargetDir C:\\output-dir -Tool devops-agent\n# Then upload each .zip from C:\\output-dir\\devops-agent-skills\\ via the Operator Web App\n```\n\n\u003C\u002Fdetails>\n\n---\n\n## ⚙️ How it works\n\n```mermaid\ngraph LR\n    S[skills\u002F] --> A[adapters\u002F]\n    ST[steering\u002F] --> A\n    A --> K[Kiro]\n    A --> CC[Claude Code]\n    A --> CU[Cursor]\n    A --> CO[Codex]\n    A --> W[Windsurf]\n    A --> GH[GitHub Copilot]\n    A --> G[Gemini CLI]\n    A --> AG[Antigravity]\n    A --> J[Junie]\n    A --> AM[Amp]\n    A --> CL[Cline]\n    A --> DA[DevOps Agent]\n```\n\n| Component | What it does |\n| --------- | ------------ |\n| **Skills** (`skills\u002F*\u002FSKILL.md`) | Self-contained, tool-agnostic playbooks. Any AI agent can follow them step-by-step. They don't depend on steering or on each other. |\n| **Steering** (`steering\u002F*.md`) | Always-on context loaded into every Kiro conversation. Other tools use equivalent mechanisms via adapters. |\n| **Powers** (`powers\u002F*\u002F`) | Bundled, installable units for Kiro. Package steering + MCP tools + hooks into a single activatable power. |\n| **Adapters** (`adapters\u002F`) | Translate steering into each tool's native config format and wire up skills as commands or rules. |\n| **Assets** (`assets\u002F`) | Shared reference material (v13 best practices, metrics, patterns) bundled with skills for tools that support it. |\n\n### Tool compatibility matrix\n\n| Tool | Steering mechanism | Skills mechanism |\n| ---- | ------------------ | ---------------- |\n| Kiro | `.kiro\u002Fsteering\u002F*.md` | `.kiro\u002Fskills\u002F*\u002FSKILL.md` |\n| Claude Code | `CLAUDE.md` | `.claude\u002Fcommands\u002F*.md` (slash commands) |\n| Cursor | `.cursor\u002Frules\u002F*.md` | Rules with conditional activation |\n| Codex | `AGENTS.md` | References `skills\u002F` directory |\n| Windsurf | `.windsurfrules` | References `skills\u002F` directory |\n| GitHub Copilot | `.github\u002Fcopilot-instructions.md` | Inline (no separate skill mechanism) |\n| Cline | `.clinerules` | References `skills\u002F` directory |\n| Gemini CLI | `GEMINI.md` | References `skills\u002F` directory |\n| Antigravity | `.agents\u002Frules\u002F*.md` | `.agents\u002Fskills\u002F*\u002FSKILL.md` |\n| Junie | `.junie\u002Fguidelines\u002F*.md` | `.junie\u002Fskills\u002F*\u002FSKILL.md` |\n| Amp | `AGENTS.md` | `.agents\u002Fskills\u002F*\u002FSKILL.md` |\n| AWS DevOps Agent | N\u002FA (skills are self-contained) | `SKILL.md` zip upload to Agent Space |\n\n### Kiro Powers\n\n[Kiro Powers](https:\u002F\u002Fkiro.dev) are bundled, installable units that activate contextually based on conversation keywords. Unlike always-on steering, Powers load guidance dynamically when relevant topics arise.\n\n| Power | Status | Description |\n| ----- | ------ | ----------- |\n| `well-architected` | 🚧 In progress | Full WA framework — all 6 pillars, activated on architecture discussions |\n\n> [!NOTE]\n> Powers are the recommended way to use WA guidance in Kiro going forward. They provide richer activation (keyword-based), optional MCP tool integration, and one-click installation from the Powers gallery.\n\n---\n\n## 📋 Skills overview\n\n| Skill | Pillar(s) | Use when you need to... |\n| ----- | --------- | ----------------------- |\n| `wa-review` | All 6 | Run a full Well-Architected review |\n| `security-assessment` | 🔒 Security | Assess IAM, detection, data protection, incident response |\n| `reliability-improvement-plan` | 🔄 Reliability | Find and eliminate single points of failure |\n| `cost-optimization-audit` | 💰 Cost Optimization | Identify waste and right-sizing opportunities |\n| `performance-efficiency` | ⚡ Performance Efficiency | Evaluate resource selection, scaling, and caching |\n| `sustainability-optimization` | 🌱 Sustainability | Reduce carbon footprint and resource waste |\n| `operational-excellence` | 🛠️ Operational Excellence | Assess CI\u002FCD, observability, incident management |\n| `migration-readiness` | All 6 | Assess readiness to migrate a workload to AWS |\n| `architecture-decision-record` | All 6 | Document a design decision with WA pillar impact |\n\n---\n\n## ✅ Verifying it works\n\nAsk your AI coding agent:\n\n```text\nWhat Well-Architected pillars should I consider for this architecture?\n```\n\nIf configured correctly, it will reference all six pillars with specific guidance rather than giving a generic answer.\n\n> [!TIP]\n> **Claude Code users**: try `\u002Fwa-review` to invoke the full review skill as a slash command.\n>\n> **Kiro users**: the steering loads automatically — just start discussing architecture and the agent applies WA principles.\n\n---\n\n## 🧪 Evaluating skills\n\nEach skill includes structured evaluations in `skills\u002F*\u002Fevals\u002Fevals.json` following the [Agent Skills eval spec](https:\u002F\u002Fagentskills.io\u002Fskill-creation\u002Fevaluating-skills). Evals let you measure whether the skills produce better outputs than a bare agent.\n\nEach test case includes:\n\n- A realistic user prompt\n- Expected output description\n- 5-7 concrete assertions (gradable as PASS\u002FFAIL)\n\n### Automated eval runner\n\nThe `evals\u002F` directory contains an automated evaluation runner powered by **Amazon Bedrock**.\n\n**Prerequisites:**\n\n- Python 3.13+ and [uv](https:\u002F\u002Fdocs.astral.sh\u002Fuv\u002F)\n- AWS credentials configured with Bedrock access (`aws configure` or SSO)\n- Bedrock model access enabled for Claude Sonnet and Haiku in your region\n\n**Setup:**\n\n```bash\ncd evals\nuv sync\n```\n\n**Run evaluations:**\n\nmacOS \u002F Linux:\n\n```bash\n# List available skills\nuv run python run.py --list\n\n# Evaluate a single skill\nuv run python run.py --skill wa-review\n\n# Evaluate all skills\nuv run python run.py\n\n# Verbose output (shows per-assertion grades)\nuv run python run.py --skill security-assessment --verbose\n\n# Save results for historical tracking\nuv run python run.py --save\n```\n\nWindows (PowerShell):\n\n```powershell\n# List available skills\nuv run python run.py --list\n\n# Evaluate a single skill\nuv run python run.py --skill wa-review\n\n# Evaluate all skills\nuv run python run.py\n\n# Verbose output (shows per-assertion grades)\nuv run python run.py --skill security-assessment --verbose\n\n# Save results for historical tracking\nuv run python run.py --save\n```\n\n> [!NOTE]\n> On Windows, ensure your AWS credentials are configured via `aws configure` or environment variables (`AWS_ACCESS_KEY_ID`, `AWS_SECRET_ACCESS_KEY`, `AWS_SESSION_TOKEN`). If using AWS IAM Identity Center (SSO), run `aws sso login --profile your-profile` first.\n\n**How it works:**\n\n1. For each test case, generates two responses via Bedrock Converse API:\n   - **Baseline** — prompt only, no skill context\n   - **With skill** — prompt + SKILL.md injected as system context\n2. An LLM-as-judge grades each assertion as PASS\u002FFAIL against both outputs\n3. Reports a score comparison showing the skill's impact\n\n**Configuration** (`evals\u002Fconfig.yaml`):\n\n```yaml\nprovider: bedrock\nregion: us-east-1\ngeneration_model: us.anthropic.claude-sonnet-4-5-20250929-v1:0\ngrading_model: us.anthropic.claude-haiku-4-5-20251001-v1:0\n```\n\n**Estimated cost per run:**\n\n| Scope | Generation calls | Grading calls | Estimated cost |\n| ----- | ---------------- | ------------- | -------------- |\n| Single skill (3 cases) | 6 (Sonnet) | 6 (Haiku) | ~$0.10 – $0.15 |\n| All 9 skills (27 cases) | 54 (Sonnet) | 54 (Haiku) | ~$0.80 – $1.20 |\n\nCost breakdown assumes ~1K input tokens and ~2K output tokens per generation call, and ~3K input \u002F ~500 output per grading call. Actual cost depends on response length and Bedrock pricing in your region. The `max_cost_usd` setting in config.yaml acts as a soft budget guard.\n\n> [!TIP]\n> **Experiment with other models!** The eval runner works with any model available in Bedrock — try Amazon Nova, Meta Llama, Mistral, or others to see how different foundation models respond to skill guidance. Use the discovery utility to see what's available in your region:\n>\n> `uv run python list_models.py`\n>\n> Then update `generation_model` in `config.yaml` to try a different model. The grading model should remain a strong model (Claude Sonnet\u002FHaiku) for reliable assertion grading.\n\n> [!TIP]\n> Start by running a single skill eval (`--skill wa-review --verbose`) to see detailed per-assertion grading. The delta between baseline and with-skill scores quantifies the value each skill adds.\n\n---\n\n## 🤝 Contributing\n\nWe welcome contributions from the community! See [CONTRIBUTING.md](CONTRIBUTING.md) for guidelines on adding skills, modifying steering files, or adding new tool adapters.\n\n> [!NOTE]\n> This is a community-driven project. Anyone can collaborate and improve the skills and steering docs through Pull Requests. Adapt them to your domain, add new patterns, and share back.\n\n---\n\n## 🔒 Security\n\nSee [CONTRIBUTING](CONTRIBUTING.md#security-issue-notifications) for more information.\n\n---\n\n## 📄 License\n\nThis project is licensed under the [MIT-0 License](LICENSE).\n\n---\n\n## 📚 Related Resources\n\n- [AWS Well-Architected Framework](https:\u002F\u002Fdocs.aws.amazon.com\u002Fwellarchitected\u002Flatest\u002Fframework\u002Fwelcome.html)\n- [Kiro — AI-powered IDE](https:\u002F\u002Fkiro.dev)\n- [AWS DevOps Agent](https:\u002F\u002Fdocs.aws.amazon.com\u002Fdevopsagent\u002Flatest\u002Fuserguide\u002F)\n- [Agent Skills Specification](https:\u002F\u002Fagentskills.io\u002F)\n- [skills.sh — Skills directory for AI agents](https:\u002F\u002Fskills.sh)\n","该项目旨在通过可重用的技能和指导，教会AI编码助手如何应用AWS Well-Architected框架。其核心功能包括提供一套涵盖12种工具（如Kiro、Claude Code、GitHub Copilot等）的统一指导手册，确保在开发过程中直接嵌入最佳实践。技术特点在于无需AWS凭证或API调用即可本地运行，并遵循开放的Agent Skills规范。适合于希望在IDE中实时获得架构设计建议，减少因不符合最佳实践而返工的开发团队使用。",2,"2026-06-11 03:58:24","CREATED_QUERY"]