multi▌

Build optimized, secure multi-stage Dockerfiles for any language or framework. \n \n Structures builds with separate builder and runtime stages, copying only necessary artifacts to minimize final image size and attack surface \n Emphasizes layer caching optimization by ordering commands from least to most frequently changing, combined with .dockerignore and command consolidation \n Recommends minimal base images (Alpine, distroless, or official slim variants) with exact version pinning for repro

Multi-agent architectures distribute work across multiple language model instances, each with its own context window. When designed well, this distribution enables capabilities beyond single-agent limits. When designed poorly, it introduces coordination overhead that negates benefits. The critical insight is that sub-agents exist primarily to isolate context, not to anthropomorphize role division.

When a company has multiple domains (e.g., company.com and product.ai), ensure the company/main site ranks first for brand queries. Product sites focus on product keywords and do not compete for brand position. See domain-architecture for structure decisions; rebranding-strategy for domain change and migration.

Intelligent multi-engine search with automatic network detection and fallback prioritization. \n \n Supports four search engines with two priority modes: quality-first (Tavily > DuckDuckGo > Bing API > Bing scraper) or balanced/free-first (DuckDuckGo > Tavily > Bing API > Bing scraper) \n Includes automatic quota management for API-based engines, 5-minute network detection caching, and forced re-detection after network changes \n Provides web content fetching and batch enrichment of search resul

Structured peer-review process using constrained agents to validate designs and surface hidden assumptions before implementation. \n \n Five specialized agent roles with hard scope limits: Primary Designer, Skeptic/Challenger, Constraint Guardian, User Advocate, and Integrator/Arbiter, each with explicit permissions and restrictions \n Sequential review workflow where the designer proposes, then three reviewer agents provide feedback in order, with mandatory Decision Log recording all objections

Access 17 search engines (8 Chinese, 9 global) without API keys for web crawling and knowledge queries. \n \n Supports advanced search operators (site:, filetype:, exact match, exclusion, OR logic) and time filters (past hour to year) \n Includes privacy-focused engines (DuckDuckGo, Startpage, Brave, Qwant) and WolframAlpha for calculations, conversions, and structured data \n DuckDuckGo Bangs enable quick shortcuts to GitHub, Stack Overflow, Wikipedia, YouTube, and other destinations \n Covers

Provides advanced multi-agent orchestration expertise for managing complex coordination of agents across distributed systems. Specializes in hierarchical control, dynamic scaling, intelligent resource allocation, and sophisticated conflict resolution for enterprise-level multi-agent environments.

Coordinate and integrate multiple omics datasets for comprehensive systems biology analysis. Orchestrates specialized ToolUniverse skills to perform cross-omics correlation, multi-omics clustering, pathway-level integration, and unified interpretation.

Decision framework and patterns for architecting applications across AWS, Azure, and GCP.

Build and deploy the same feature consistently across web, mobile, and desktop platforms using API-first architecture and parallel implementation strategies.