> skills // review & truth
brainstrust
A second opinion grounded in your actual codebase — the model reads the repo itself.
// what it is
Instead of hand-feeding files to a reviewer, brainstrust lets the consulted model explore your repo on its own with read, grep and find. Claude consults ride your subscription through subagents at no extra cost; non-Anthropic voices go through an OpenRouter agent loop, so you get a genuinely diverse panel.
GOOD FOR // A grounded review, an architecture sounding board, or a devil’s advocate before you commit to a direction.
// how it works
Built-in methodologies: review, architecture, debug, security, devil’s advocate, strategy, explore, ideate
The panel is input, not verdict: you synthesise and decide
// install
It is a Claude Code plugin. Add the marketplace, then install. Two lines in the Claude Code prompt:
/plugin marketplace add jezweb/brainstrust
/plugin install brainstrustalso runs as a CLI: npx tsx src/consult.ts
source & docs: github.com/jezweb/brainstrust