What problem does it solve? Studying an unfamiliar open-source codebase by hand is slow and shallow: you skim the README, miss the architecture, and never surface the anti-patterns worth avoiding. This Skill runs a structured multi-agent research pipeline that investigates, deepens, inverts, and synthesizes findings from any external Git repository. ## Core Features & Use Cases - 7-Phase Research Pipeline: Clones a repository (with HTTPS-only host allowlisting and HEAD SHA pinning), then runs parallel Explore and general-purpose agents to map architecture, deepen key areas, invert assumptions, and hunt for blunders. - Multi-Model Synthesis: Spawns parallel Opus and Sonnet Plan agents that write application-focused and ergonomics-focused findings to docs/research/, then merges everything into a sourced research proposal. - Major Feature Integration Mode: Extends research with an integration proposal, iterative deepening, five blunder-hunt passes, and cross-model feedback when the goal is adopting a feature rather than just learning. - Use Case: Point it at a repo like a terminal multiplexer or agent framework you admire, and receive a structured proposal at docs/research-<repo>-<date>.md covering what to adopt, what to avoid, and how it maps to your codebase. ## Quick Start Research the GitHub repository at https://github.com/example/project and produce a proposal of insights we can apply to our codebase.
