一键导入
setup
Configure which review agents run for your project. Auto-detects stack and writes compound-engineering.local.md.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Configure which review agents run for your project. Auto-detects stack and writes compound-engineering.local.md.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Guide for structuring, parsing, and validating XML-like tags in LLM inputs, system prompts, outputs, and agent-to-agent communication. Use this skill when designing prompt templates, parsing structured content from text, or handling nested tag boundaries.
Implementation reference library for compound engineering workflows. For workflow orchestration, use slash commands under `commands/workflows/` (constitution, plan, work, review, compound). This skill provides stack-specific fix recipes those workflows consume.
Browser automation using Vercel's agent-browser CLI. Use when you need to interact with web pages, fill forms, take screenshots, or scrape data. Alternative to Playwright MCP - uses Bash commands with ref-based element selection. Triggers on "browse website", "fill form", "click button", "take screenshot", "scrape page", "web automation".
Build applications where agents are first-class citizens. Use this skill when designing autonomous agents, creating MCP tools, implementing self-modifying systems, or defining architecture artifacts around deletion tests, interfaces, seams, adapters, and explicit contracts.
Run a comprehensive scored audit of agent-native architecture principles
This skill should be used before implementing features, building components, or making changes. It guides exploring user intent, approaches, and design decisions before planning. Triggers on "let's brainstorm", "help me think through", "what should we build", "explore approaches", ambiguous feature requests, or when the user's request has multiple valid interpretations that need clarification.
| name | setup |
| description | Configure which review agents run for your project. Auto-detects stack and writes compound-engineering.local.md. |
| model | claude-sonnet-4.6 |
| platforms | {"claude":{"disable-model-invocation":true},"copilot":{"model":"gpt-5.3-codex"},"opencode":{"model":"openrouter/moonshotai/kimi-k2.6"}} |
Interactive setup for compound-engineering.local.md -- configures which agents run during /workflows:review and /workflows:work, plus the visible Ralph/TDD defaults that planning and execution inherit.
Read compound-engineering.local.md in the project root. If it exists, display current settings summary (including the TDD contract) and use AskUserQuestion:
question: "Settings file already exists. What would you like to do?"
header: "Config"
options:
- label: "Reconfigure"
description: "Run the interactive setup again from scratch"
- label: "View current"
description: "Show the file contents, then stop"
- label: "Cancel"
description: "Keep current settings"
If "View current": read and display the file, then stop. If "Cancel": stop.
Auto-detect the project stack by checking for common project indicators:
# Detect project stack from root files
if [ -f "artisan" ] && [ -f "composer.json" ]; then
echo "laravel"
elif [ -f "Cargo.toml" ]; then
echo "rust"
elif [ -f "nest-cli.json" ] || grep -q '"@nestjs/core"' package.json 2>/dev/null; then
echo "nestjs"
elif [ -f "nuxt.config.ts" ] || [ -f "nuxt.config.js" ]; then
echo "nuxt"
elif [ -f "next.config.ts" ] || [ -f "next.config.js" ] || [ -f "next.config.mjs" ]; then
echo "nextjs"
elif [ -f "angular.json" ]; then
echo "angular"
elif [ -f "vite.config.ts" ] && grep -q '"vue"' package.json 2>/dev/null; then
echo "vue"
elif [ -f "vite.config.ts" ] && grep -q '"react"' package.json 2>/dev/null; then
echo "react"
elif [ -f "composer.json" ]; then
echo "php"
elif [ -f "tsconfig.json" ]; then
echo "typescript"
elif [ -f "package.json" ]; then
echo "javascript"
elif [ -f "pyproject.toml" ] || [ -f "requirements.txt" ] || [ -f "setup.py" ]; then
echo "python"
elif [ -f "go.mod" ]; then
echo "go"
else
echo "general"
fi
Use AskUserQuestion:
question: "Detected {type} project. How would you like to configure?"
header: "Setup"
options:
- label: "Auto-configure (Recommended)"
description: "Use smart defaults for {type}. Done in one click."
- label: "Customize"
description: "Choose stack, focus areas, and review depth."
[laravel-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][laravel-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][vue-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][typescript-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][typescript-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][nestjs-reviewer, typescript-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][typescript-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][typescript-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][python-reviewer, code-simplicity-reviewer, security-sentinel, performance-oracle][code-simplicity-reviewer, security-sentinel, performance-oracle, architecture-strategist][code-simplicity-reviewer, security-sentinel, performance-oracle, architecture-strategist][code-simplicity-reviewer, security-sentinel, performance-oracle, architecture-strategist]a. Stack -- confirm or override:
question: "Which stack should we optimize for?"
header: "Stack"
options:
- label: "{detected_type} (Recommended)"
description: "Auto-detected from project files"
- label: "Laravel"
description: "Laravel PHP -- adds Laravel convention reviewer"
- label: "Vue/Nuxt"
description: "Vue.js / Nuxt -- adds frontend reviewer"
- label: "React/Next.js"
description: "React / Next.js -- adds TypeScript reviewer"
- label: "Angular"
description: "Angular -- adds TypeScript reviewer"
- label: "NestJS"
description: "NestJS -- adds NestJS and TypeScript reviewers"
- label: "Python"
description: "Python -- adds Pythonic pattern reviewer"
- label: "Rust"
description: "Rust -- adds architecture reviewer"
- label: "Go"
description: "Go -- adds architecture reviewer"
Only show options that differ from the detected type.
b. Focus areas -- multiSelect:
question: "Which review areas matter most?"
header: "Focus"
multiSelect: true
options:
- label: "Security"
description: "Vulnerability scanning, auth, input validation (security-sentinel)"
- label: "Performance"
description: "N+1 queries, memory leaks, complexity (performance-oracle)"
- label: "Architecture"
description: "Design patterns, SOLID, separation of concerns (architecture-strategist)"
- label: "Code simplicity"
description: "Over-engineering, YAGNI violations (code-simplicity-reviewer)"
c. Depth:
question: "How thorough should reviews be?"
header: "Depth"
options:
- label: "Thorough (Recommended)"
description: "Stack reviewers + all selected focus agents."
- label: "Fast"
description: "Stack reviewers + code simplicity only. Less context, quicker."
- label: "Comprehensive"
description: "All above + git history, data integrity, agent-native checks."
d. Delivery loop:
question: "What should the default delivery loop be for this repo?"
header: "TDD"
options:
- label: "Ralph-driven TDD (Recommended)"
description: "Failing tests first, minimal implementation second, refactor third, rerun until green and clean. Plans default to unit + e2e evidence unless they record a justified exception."
- label: "Standard implementation"
description: "Implementation can lead. Plans may still opt into Ralph explicitly, but any weaker test evidence should be recorded in the plan."
e. Review mode:
question: "When should reviews run during /workflows:work?"
header: "Review mode"
options:
- label: "Bulk (Default)"
description: "Review runs once at the end of all tasks. Faster, less interruption."
- label: "Inline"
description: "Review runs after each task. Slower, but catches issues earlier."
- label: "Both"
description: "Inline review per task AND bulk review at the end."
Stack-specific agents:
laravel-reviewervue-reviewertypescript-reviewernestjs-reviewer, typescript-reviewerpython-reviewerFocus area agents:
security-sentinelperformance-oraclearchitecture-strategistcode-simplicity-reviewerDepth:
code-simplicity-reviewer onlygit-history-analyzer, data-integrity-guardian, agent-native-reviewerPlan review agents: stack-specific reviewer + code-simplicity-reviewer.
Execution settings:
tdd_enabled: compatibility mirror for current execution templates. true when tdd.mode is ralph, otherwise false.tdd.precedence: plan_overrides_local (any concrete plan tdd value beats this file; inherit falls back here)tdd.mode: ralph or standardtdd.loop: red-green-refactor or implementation-firsttdd.evidence.unit: required or optionaltdd.evidence.e2e: required or optionaltdd.exceptions: [] by default. Plans must carry any justified exceptions.review_mode: "bulk" (default), "inline", or "both" (controls per-task review in workflows:work)Write compound-engineering.local.md:
---
review_agents: [{computed agent list}]
plan_review_agents: [{computed plan agent list}]
tdd_enabled: {true|false}
tdd:
precedence: plan_overrides_local
mode: {ralph|standard}
loop: {red-green-refactor|implementation-first}
evidence:
unit: {required|optional}
e2e: {required|optional}
exceptions: []
review_mode: bulk
---
# Review Context
Add project-specific review instructions here.
These notes are passed to all review agents during /workflows:review and /workflows:work.
## TDD Defaults
- `tdd` is the visible repo-local default contract for planning and execution.
- Plan-level `tdd` values override this file for that plan; `inherit` falls back to these defaults.
- `tdd_enabled` mirrors whether `tdd.mode` is `ralph` until execution templates read the full `tdd` block directly.
Examples:
- "We use Turbo Frames heavily -- check for frame-busting issues"
- "Our API is public -- extra scrutiny on input validation"
- "Performance-critical: we serve 10k req/s on this endpoint"
Saved to compound-engineering.local.md
Stack: {type}
Review depth: {depth}
Agents: {count} configured
{agent list, one per line}
TDD default: {ralph|standard}
Loop: {red-green-refactor|implementation-first}
Evidence: unit={required|optional}, e2e={required|optional}
Precedence: plan `tdd` values override local defaults when set
Tip: Edit the "Review Context" section to add project-specific instructions.
Re-run this setup anytime to reconfigure review agents or TDD defaults.