Model Crosscheck — AI skill for Claude Code
Adversarial multi-model planning & code review for Claude Code: /dual-plan and /dual-review (Claude + OpenAI Codex), /tri-plan and /tri-review (adds Google Gemini).
How to install Model Crosscheck
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open SameerKhan/model-crosscheck and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Model Crosscheck does
Adversarial multi-model planning & code review for Claude Code: /dual-plan and /dual-review (Claude + OpenAI Codex), /tri-plan and /tri-review (adds Google Gemini). Independent models critique the same plan or diff; agreement is signal, disagreement is a map.
Alternatives in AI
- Codex — openai-codex Adversarial code review, Codex CLI integration, cross-model analysis 9.6k ★
- Fusion Plan — OMC interview (interactive) → 3-round Opus 4.8 + GPT-5.5 fusion deepening → concise .omc/plans/ output → revie 468 ★
- Code CLI — Open-source terminal coding agent 205 ★
README
model-crosscheck: two (or three) models, checking each other
**Renamed from `dual-ai-skills`.** Installed before 2026-07-25? Run `/plugin marketplace update dual-ai-skills`; the plugin is recognized as renamed and keeps working, but the skill namespace becomes `/crosscheck:*`. See [Renamed from dual-ai-skills](#renamed-from-dual-ai-skills).
[Claude Code](https://claude.com/claude-code) skills that put **Claude, OpenAI's Codex CLI, and, for the five `tri-*` skills, Google's Gemini, in an adversarial loop** on the same work, instead of trusting any one model alone.
Which one do I want?
| You are about to… | Use | You get |
|---|---|---|
| write code from a plan | /dual-plan · /tri-plan |
a plan rival models signed off on, before a line is written |
| merge a diff | /dual-review · /tri-review |
findings ranked by how many independent models found them |
| pick a datastore, framework, protocol | /tri-decide |
a decision record: options, bets, cruxes, revisit trigger |
| price, package, or bet the business | /tri-strategy |
a strategy memo argued from three assigned lenses |
| rely on an outside fact | /tri-research |
a ledger where every claim is re-fetched and status-marked |
Rule of thumb: **`dual` = Claude + Codex. `tri` = both plus Gemini.** More models is not automatically better, see [When *not* to use these](#when-not-to-use-these).
The skills
On code
/dual-plan: Claude drafts an implementation plan, Codex critiques it against the actual codebase (read-only), Claude adjudicates each point, and they loop until Codex signs off SOUND (max 3 rounds; after that, unresolved disagreements are surfaced to you as a decision point, never papered over). You approve the converged plan before any code is written./dual-review: before a merge, Claude's/code-reviewandcodex exec reviewrun in parallel on the same diff. The findings are merged: agreement between two different models is a strong s
Related Skills
Santa Review
Adversarial dual-review convergence loop — two independent model reviewers must both approve before code ships
V Review Plan
Run an independent cross-model (Codex) adversarial review of a Compound V plan/manifest before dispatch, then
Dual Review File Based
File-based cross-review and correction between independent AI agents / Файловое кросс-ревью и исправление межд
Maestro Skill
Multi-model build loop for Claude Code, Codex, Cursor and agy: the session model conducts, models from other f
Claude GPT Workflow
Claude Code skills for Claude + GPT collaborative workflows. Adversarial plan review + orchestrated code execu
Codex Plan Review
Adversarial Codex review of a PLAN or design (not code) via the local Codex CLI. Read-only second-model challe
Related Agents
Deep Repo Auditor
Cross-validated deep repository audit — dual AI models (Sonnet + Codex) in parallel, consolidated report with
CLI Subagents Antigravity
Runs a task in Google Antigravity (agy, Gemini models) on the user's Google account (CLI Subagents; extra Goog
Claude Spec Review
First-pass adversarial spec review by Claude. Read-only and advisory. Runs before the Codex spec-reviewer loop