deepstack（deepstack 搜索）

Deep Stack Research Prompt Generator

Deep stack analysis tool.

Arguments: $ARGUMENTS

Instructions

Generate a comprehensive deep research prompt for the topic "$ARGUMENTS" tailored to the current project's technology stack.

If no topic is provided, ask the user what topic they want to research (e.g., security, performance, maintainability, scalability, testing, deployment, accessibility).

Phase 1: Detect the Technology Stack

Analyze the current project to identify all technologies and their versions. Check these files:

Backend Detection

Frontend Detection

Check package.json dependencies for:

• Frameworks: React, Vue, Svelte, Angular, Solid, Preact, Qwik
• Meta-frameworks: Next.js, Nuxt, SvelteKit, Remix, Astro
• Build tools: Vite, Webpack, Parcel, esbuild, Turbopack
• CSS: Tailwind CSS, Bootstrap, styled-components, Sass, PostCSS

Integration/Middleware Layer

Look for glue technologies:

• Inertia.js (@inertiajs/vue3, @inertiajs/react, etc.)
• Livewire (in composer.json)
• HTMX (in package.json or HTML files)
• Turbo/Hotwire (in package.json or Gemfile)
• Alpine.js (often paired with Livewire/HTMX)

Database Detection

Check these locations:

• .env file: DB_CONNECTION, DATABASE_URL
• docker-compose.yml: Database service definitions
• config/database.php (Laravel), settings.py (Django), etc.
• ORM configs: Prisma (schema.prisma), Drizzle, TypeORM, Eloquent

Common databases: MySQL, PostgreSQL, SQLite, MongoDB, Redis, Elasticsearch

Infrastructure/Other Tools

• Caching: Redis, Memcached
• Search: Elasticsearch, Meilisearch, Algolia
• Queues: Redis, RabbitMQ, SQS
• Storage: S3, local filesystem
• Auth: Sanctum, Passport, NextAuth, Auth0, Clerk

Version Detection

For each technology found, extract the version:

• package.json: Check dependencies and devDependencies for exact versions
• composer.json: Check require section
• Lock files (package-lock.json, composer.lock) have exact versions
• .env or Docker configs may specify database versions

Phase 2: Confirm Stack with User

IMPORTANT: First output the detected stack as a text message to the user. Do NOT put the stack details inside the AskUserQuestion tool — the tool has limited space for display.

Step 1: Display the stack as text output

Output a message like this (with actual detected values):

I detected the following technology stack for this project:

**Backend:**
- Rust 1.75
- Tauri 2.0

**Frontend:**
- TypeScript 5.3
- React 18.2
- Vite 5.0
- Tailwind CSS 3.4

**Database:**
- SQLite (via Tauri)

**Desktop:**
- Tauri 2.0 (Rust backend, webview frontend)

**Other:**
- pnpm (package manager)

Step 2: Then ask for confirmation

AFTER displaying the stack details above, use the AskUserQuestion tool with a simple confirmation question:

• Question: "Does this detected stack look correct?"
• Options: "Yes, looks correct" / "Need corrections"

Wait for user confirmation before proceeding. If the user provides corrections, incorporate them into the stack.

Phase 3: Generate the Deep Research Prompt

Once the stack is confirmed, generate a comprehensive research prompt. The prompt should be structured for a deep research tool (like Claude, Perplexity, or similar).

Prompt Template

Generate output in this format (replace placeholders with actual stack details):

START OF GENERATED PROMPT

I'm working on a web application with the following technology stack:

Backend:

• [List each backend technology with version]

Frontend:

• [List each frontend technology with version]

Integration:

• [List integration/middleware if applicable]

Database:

• [List databases and data stores with versions]

Other:

• [List other notable tools]

I need comprehensive research on [TOPIC from $ARGUMENTS] best practices, patterns, and considerations for this specific stack.

Research Scope

1. Individual Technology Analysis

For each technology in my stack, research:

• [TOPIC] best practices specific to this version
• Known issues, vulnerabilities, or limitations related to [TOPIC]
• Configuration options that affect [TOPIC]
• Common mistakes developers make regarding [TOPIC]
• Version-specific considerations (what changed in recent versions)

2. Integration Points

Research [TOPIC] considerations for these technology combinations:

• [Backend framework] + [Frontend framework] (data flow, state management)
• [Backend] + [Database] (query patterns, connection handling)
• [Integration layer] specifics (if applicable)
• [Any other relevant combinations based on the stack]

Focus on issues that arise specifically from these technologies working together, not just individual concerns.

3. Stack-Specific Patterns

Identify [TOPIC] patterns and architectures that are:

• Recommended for this exact stack combination
• Anti-patterns to avoid with this stack
• Trade-offs specific to these technology choices

4. Real-World Considerations

Research:

• Common [TOPIC] issues reported by developers using this stack
• Production lessons learned
• Scaling considerations related to [TOPIC]
• Monitoring and observability for [TOPIC]

Requested Output Format

Please provide your findings organized as:

1. Executive Summary

- Top 10 most critical [TOPIC] considerations for this stack - Priority-ranked action items

1. Per-Technology Guidelines

- Organized by each technology in the stack - Specific, actionable recommendations - Code examples where helpful

1. Integration Guidelines

- [TOPIC] at the boundaries between technologies - Data flow considerations - Common pitfalls when technologies interact

1. Known Issues & Gotchas

- Version-specific bugs or limitations - Documented vulnerabilities (for security topics) - Edge cases to watch for

1. Checklist

- Actionable audit checklist for [TOPIC] - Can be used to review existing code

1. Anti-Patterns

- What NOT to do - Common mistakes with this stack - Why they're problematic

1. Resources

- Official documentation links - Recommended articles, tutorials, talks - Tools that help with [TOPIC] for this stack

END OF GENERATED PROMPT

Phase 4: Present the Output

Output the generated prompt as plain text that the user can easily copy.

Before the prompt, add:

Here's your deep research prompt for "[TOPIC]" tailored to your stack. Copy this and paste it into your preferred research tool:

After the prompt, add:

Tip: This prompt works well with Claude, ChatGPT, Perplexity, or similar AI research tools. For best results, use a tool that can search the web for current information.

Topic-Specific Additions

Depending on the topic provided in $ARGUMENTS, emphasize different aspects:

If topic is "security":

• Emphasize CVEs, OWASP Top 10, authentication, authorization, input validation
• Include encryption, secrets management, dependency vulnerabilities
• Request exploit examples and mitigation strategies

If topic is "performance":

• Emphasize profiling, caching strategies, database optimization, lazy loading
• Include bundle size, Core Web Vitals, memory management
• Request benchmarking approaches and monitoring tools

If topic is "testing":

• Emphasize unit, integration, e2e testing strategies for the stack
• Include mocking strategies, test data management, CI/CD integration
• Request coverage goals and testing anti-patterns

If topic is "maintainability":

• Emphasize code organization, naming conventions, documentation
• Include refactoring patterns, technical debt management
• Request code review checklists and architecture patterns

If topic is "scalability":

• Emphasize horizontal/vertical scaling, load balancing, caching
• Include database sharding, microservices considerations
• Request capacity planning and bottleneck identification

If topic is "deployment":

• Emphasize CI/CD, containerization, environment management
• Include rollback strategies, zero-downtime deployments
• Request infrastructure as code and monitoring setup

If topic is "accessibility":

• Emphasize WCAG compliance, screen reader support, keyboard navigation
• Include ARIA patterns, color contrast, focus management
• Request testing tools and audit approaches

Notes

• Always include version numbers — they matter for accurate research
• If you can't detect a version, note it as "[version unknown]" and ask the user
• The generated prompt should be self-contained and not require additional context
• Tailor the integration section to the actual technologies detected (don't include generic examples)