Cross Eval

Looking at what's here, this appears to be a cross-evaluation tool, likely for comparing outputs or assessing results across different models, runs, or criteria. The source file doesn't give me implementation details to work with, so I can't tell you exactly how it structures comparisons or what format it expects. If you're running multiple AI agents or model variations and need to systematically compare their outputs, this is probably the kind of utility you'd reach for. Without seeing the actual mechanics, I'd say proceed with caution and check the implementation details before building any critical workflows around it. Could be genuinely useful for A/B testing scenarios if it does what the name suggests.

skills.sh installs: 376
GitHub stars: 29.5K
last updated: Aug 21, 2026

View on GitHub

Install

$ npx -y skills add alirezarezvani/claude-skills --skill cross-eval --agent claude-code

Installs into .claude/skills of the current project.

Files

/c-level-agents:cross-eval — Multi-Model Consensus

Command: /c-level-agents:cross-eval <memo-or-brief>

Runs the same memo through multiple model providers and reconciles divergences. Use for high-stakes, irreversible decisions where single-model bias is too costly: M&A, major fundraises, layoffs, strategic pivots, regulatory commitments.

Adapted from gstack's /codex cross-review pattern, generalized to business memos instead of code PRs.

When to Run

Before signing a term sheet
Before announcing a layoff
Before committing to a regulated market
Before any decision where reversing costs > 6 months of company time
When the boardroom vote was split or had a CRITICAL dissent

Models Used (graceful degradation)

The command tries to invoke each available model in order:

Claude (primary, always available) — the boardroom's native voice
Codex / OpenAI (if OPENAI_API_KEY or codex CLI available)
Gemini (if GEMINI_API_KEY or gemini CLI available)

If only Claude is available, the command runs Claude-only with adversarial mode — same model, different prompt seeds — and clearly labels the output as single-model.

SKILL.md on GitHub