dependency-architecture

Routes engineering work to the smallest suitable skill flow and checks prerequisites.

Uses Serena MCP for symbol discovery, references, dependency understanding, and precise symbol-level edits.

Uses ast-grep for AST-aware code search, lint rules, and controlled rewrites while keeping text search and semantic navigation separate.

Initializes a repository for portable agent workflows without overwriting existing conventions.

Build token-budgeted repository maps and route large-code discovery between Serena, Repomix, and Probe.

Audit agent skills, MCP servers, plugins, hooks, prompts, permissions, pins, licenses, and content digests before activation.

Verify implementation claims with relevant tests, validation, diffs, and concrete evidence before completion.

Capture browser performance traces and turn Core Web Vitals, network, console, and rendering evidence into actionable reports.

Run token-efficient browser quality assurance with Playwright CLI and escalate to MCP only for persistent exploratory loops.

Retrieve current version-specific library and API documentation from Context7 or primary official sources.

Convert engineering findings into concise terminal, JSON, Markdown, SARIF, HTML, and Mermaid-backed reports.