presearch

To generate a technical brief, invoke presearch with a topic to produce a structured briefing containing concrete API shapes, exact package versions, data models, dependency lists, and environment variables for Gemini synthesis.

Stack-aware architecture planning detects existing project markers like package.json and CLAUDE.md to constrain technical recommendations to the current tech stack, ensuring generated briefings match your repository's established dependencies.

You can refine technical research briefings using optional modes such as --deep for comprehensive analysis, --quick for faster results, and --no-ship, alongside iterative Q&A and web research to produce an actionable plan.

The best way to define SDKs and architecture patterns is generating a structured technical briefing that includes a greenfield or project structure outline, exact package names, and dependency lists tailored for Gemini synthesis.

Yes, presearch works with existing repositories by scanning for markers like package.json and CLAUDE.md to seed constraints, automatically tailoring API, SDK, and dependency recommendations to your current project stack.

You should avoid automated technical research when your project lacks defined markers like package.json or CLAUDE.md for stack constraints, or when you need immediate code generation rather than a comprehensive pre-ship structured briefing.