interns-review-plugin
A Claude Code plugin that automatically spawns skeptical reviewer subagents to poke holes in every solution your coding agent proposes, before you implement it.
🔗 Visit interns-review-plugin
Description
AI coding agents are good at proposing a plausible-looking solution quickly, but plausible isn't the same as correct — edge cases get missed, simpler approaches get overlooked, and a confident-sounding agent doesn't flag its own blind spots. This plugin adds a check on that by inserting an adversarial review step: whenever your coding agent proposes a solution (automatically, or on-demand via a /consensus command), it spawns 1-3 independent reviewer subagents whose job is specifically to critique the proposal, not rubber-stamp it.
Each reviewer looks through a different lens — correctness, edge cases, unnecessary complexity — and the lead agent then verifies their findings against the actual code before accepting or rejecting them, rather than blindly trusting every critique. The output is a compact receipt showing what was accepted or rejected and why, tagged by severity, so you get a quick audit trail instead of having to read three separate review essays. It's a free, open-source Claude Code plugin.
💬 Our review
The short version: this is a lightweight, sensible pattern — using independent adversarial subagents to critique an agent's own proposal is a known technique for catching mistakes a single pass misses, and packaging it as a drop-in Claude Code plugin makes it accessible without building your own multi-agent review harness.
It occupies a specific niche: not a general code review tool for human PRs (like a GitHub Action linter or a human-facing review bot), but specifically a pre-implementation gate on an AI agent's own proposed solutions, inside the Claude Code workflow. The lead-agent-verifies-findings step is a good design choice — it avoids the common failure mode where every reviewer subagent's opinion gets accepted uncritically, adding noise instead of signal. As a young, single-author open-source plugin, there's limited track record, and its usefulness scales with how much you're already leaning on Claude Code to write non-trivial code unsupervised. Worth adding if you use Claude Code for meaningful coding tasks and want a cheap safety net before code gets implemented; unnecessary if you're already reviewing every agent proposal closely yourself.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Open source, gratuit
Pros
Ajoute une revue adversariale automatique avant implémentation
Plusieurs angles de critique (correction, edge cases, simplicité)
L'agent principal vérifie les critiques avant de les accepter, filtre le bruit
Gratuit, open source, plugin Claude Code prêt à l'emploi
Cons
Projet jeune, auteur unique, peu de recul
Spécifique à Claude Code, pas un outil de revue générique pour PR humaines
Utilité proportionnelle à la confiance déjà accordée à l'agent sans supervision