What problem does it solve? Single-perspective code reviews often miss issues or reflect one reviewer's bias. This Skill convenes eight opinionated engineering personas (Beck, Fowler, Uncle Bob, DHH, Metz, Evans, Hickey, Ousterhout) to review the same code independently, then surfaces only the improvements that reach consensus across fundamentally different philosophies. ## Core Features & Use Cases - Parallel multi-persona review: Dispatches eight reviewer agents (via native subagents when available, or sequential single-session passes as fallback) that read the full codebase and return structured critiques. - Consensus synthesis: Groups semantically similar findings, counts agreement, and splits results at a 6/8 threshold into an action plan versus discussion points, while preserving singleton observations. - Scoped or full-codebase review: Defaults to reviewing all source files rather than just a diff, unless the user explicitly narrows the scope to a file, directory, or commit range. - Use Case: Before merging a large feature branch, run the panel over the repository to get a prioritized, consensus-backed list of refactoring and design improvements, then type "doe het" to execute the agreed fixes one commit at a time. ## Quick Start Ask the agent to run the software guru panel over the current repository and present the consensus action plan.
