adversarial-review

Claude Code plugin: universal radial-tree exploration engine. One tree skill + swappable presets (brainstorm / attack / design / code-audit) for divergent ideation, adversarial critique, and design-space exploration. 12 framings × hard-ban-on-incomplete-leaves × stable convergence.

两仪 Liangyi · 跨厂商多 Agent 产品想法优化工作流

Codebase-level adversarial review by a panel of frontier models. A Claude Code skill that runs every file through DeepSeek + Gemini + Kimi + MiniMax in sequence, then has Claude verify the findings against the actual source.

A virtual board of experts for decisions that are expensive to redo.

Autonomous scientific discovery engine - 82 discovery modules, 16-phase pipeline with refinement and reflexion. GFlowNet hypothesis tournaments, adversarial skeptic review, multi-agent debate, causal reasoning, knowledge graphs, 17 domain ontologies. Terminal-native.

Decision gate for AI-era work: adversarial review for code, research, and production decisions.

OmniForge. Because merge requests deserve more than LGTM. 3 adversarial AI agents tear apart your code in isolated worktrees, auto-fix the findings, and forge MRs that actually have descriptions. One plugin to forge your MR lifecycle. Zero excuses.

Adversarial AI code review that declares what the round could not close. Python CLI + GitHub Action, MIT.

Research workrooms for AI agents: evidence retrieval, citation audit, adversarial review, peer review gates, and clean final reports.

Evidence-first coding agent for Claude Code CLI. Port of burkeholland/anvil from GitHub Copilot CLI.

Cross-model delegation and adversarial review for CLI coding agents: one lead, six runtime adapters, and no change merges reviewed only by its own model family.

Master of Ceremonies (emcee) — a harness-agnostic operating regimen for AI coding agents: roles, quality gates, and an adversarial review panel across Claude Code and Codex.

Two-stage adversarial validity audit for AI-produced work: mechanical checks plus an independent cold reviewer, gate freeze, miss ledger, and golden-case recall benchmarks.

Portable Claude/Codex/WorkBuddy skill: AllinCMS content ops (raw → wiki → draft → adversarial review → publish → reflect). 6 rounds of adversarial review applied; Karpathy-style living wiki.

Multi-model consensus system — 3 LLMs cross-examine each other to catch blind spots on critical decisions

Role-based AI committee Codex Skill for complex decisions, adversarial critique, and evidence-driven verification.

A loop-engineering system for AI coding agents, delivered as a Pi package. Converge → specify → implement: parallel TDD development with dual-ring convergence gates, adversarial cross-source review, source-grounded citation verification, and prior-art collision search — honest coverage at every gate, never a silent green.

Adversarial AI review that returns a structured verdict: a score, findings classified by severity, and a clear pass/fail floor.

An adversarial AI code reviewer that attacks contracts, invariants, architecture, and bugs before they reach main.