mastra

The Mastra skill serves as an authoritative technical companion for developers building AI-powered applications using the Mastra framework. Designed to mitigate the risks of using outdated training data, this skill enforces a strict verification-first methodology, ensuring all code generation relies on the specific version of Mastra installed in the local environment. It is intended for software engineers, AI architects, and developers who require high-precision integration of LLM-based agents, complex workflows, RAG systems, and memory modules within a TypeScript ecosystem. By providing integrated tools for dependency checking, embedded documentation traversal, and remote API syncing, this skill acts as a command center for Mastra-specific development tasks, including project scaffolding, troubleshooting of common type mismatches, and migration between framework versions.

• Real-time documentation verification using local node_modules lookup to ensure compatibility with your current installation.

• Structured guidance for building autonomous Agents, pipeline-based Workflows, and Tool-use integrations.

• Advanced debugging support for TypeScript configuration (ES2022/Module requirements) and runtime error resolution.

• Model router management for provider-specific LLM implementations using standardized registry scripts.

• Automated environment setup for Mastra Studio, enabling interactive testing and performance observability.

• Migration assistance to manage breaking changes and version upgrades across Mastra core components.

• Always run local lookup commands before generating code to avoid hallucinations regarding constructor signatures.

• Use the provided provider-registry scripts to verify available models before implementing model router strings.

• Configure TSConfig with the specified ES2022 module resolution settings to prevent import/module failures.

• Prioritize the embedded doc references over external knowledge for API surface areas and type definitions.

• Input: User requirements for agent behavior, workflow logic, or RAG architecture. Output: Validated code snippets, architectural patterns, and troubleshooting steps tailored to the active version of Mastra.