← ClaudeAtlas

ai-jurylisted

Convene a cross-vendor multi-agent review jury on a diff, PR, or issue and produce one report — Claude Code, Codex, Antigravity, and/or a free local/open-weight model each review independently, cross-examine each other, verify, and reach one verdict — a chair's synthesis or a panel vote. Handles the whole flow end to end (scaffold config if needed → review → report → summarize). Use when the user wants a multi-model review of a pull request, a diff, an issue, or the current branch, or says "review jury", "convene the jury", or "cross-model review".
berkayturanci/ai-jury · ★ 3 · AI & Automation · score 69
Install: claude install-skill berkayturanci/ai-jury
# AI Jury Convene a panel of native coding-agent CLIs from **different vendors** to review the same change, debate each other's findings, and produce one consolidated verdict. This is multi-model review where each agent runs in its own native CLI (with its own tooling), not API-level prompting. ## Prerequisites - `ai-jury` is installed and `jury` is on PATH (`pipx install ai-jury` or `pip install -e /path/to/ai-jury`). - At least one reviewer: an agent CLI (`claude`, `codex`, and/or `agy`) **or** a free, offline **local / open-weight** model via Ollama (configured as a `vendor = "local"` agent). Missing/unreachable agents are skipped; the jury runs with whoever is available. - `gh` is authenticated for `--pr` / `--post`. - First time in a repo? Run `jury init` to scaffold a `jury.toml` (it detects installed agents and local models). ## How to run Pick the form that matches the request: | Intent | Command | |:--|:--| | Review a GitHub PR | `jury --pr <number>` | | Review and post the verdict as a PR comment | `jury --pr <number> --post` | | Review the current branch vs default | `git diff origin/HEAD... \| jury --diff-file -` | | Review a diff file | `jury --diff-file path/to/changes.diff` | | Single-round (no debate) | add `--rounds 1` | | Offline smoke test (no live CLIs) | `jury --mock --diff-file examples/sample.diff` | Stream progress goes to stderr; the markdown report goes to stdout (or `-o file.md`). ## Parameters (the ones you'll reach for) Grouped by