analyzing-projects

The analyzing-projects skill is a powerful diagnostic tool designed for software engineers and developers who need to quickly gain situational awareness of unfamiliar repositories. It functions as an automated expert guide that methodically audits a project's architecture, identifying core technologies, dependencies, and established conventions. By executing a structured workflow, the agent synthesizes raw directory data and configuration files into a coherent, actionable summary, significantly reducing the cognitive load required to understand how a codebase functions or where specific logic resides.

• Automatically detects and categorizes tech stacks, including package managers (npm, pip, cargo, gomod), frameworks (Next.js, FastAPI, Spring Boot), and infrastructure as code tools (Terraform, Kubernetes).

• Generates a comprehensive architectural overview, mapping source code directories, entry points, and test suites into a visual tree structure.

• Identifies key architectural patterns (monolith, microservices, serverless), API styles (REST, GraphQL, gRPC), and database choices.

• Evaluates development workflows by inspecting linters, formatters, Git hooks (Husky), and build scripts (Makefile, npm scripts).

• Produces a standardized, easy-to-read markdown summary covering architecture, directory purposes, entry points, development commands, and project-specific coding conventions.

• Provides a robust validation checklist to ensure no critical components are overlooked during the investigation.

• Ideal for use when starting on a new project, performing code audits, or whenever you need a quick answer to "how does this work?" or "what is the architecture?".

• Input is typically a local directory path, and output is a structured summary report.

• Can be used in conjunction with other workflow plugins or agents to jumpstart development, debugging, or documentation tasks.

• Operates best when provided with read access to the project root and its configuration manifests.

• Note that while the tool is automated, it performs best when the codebase maintains standard conventions; non-standard structures may require manual verification of the suggested report.