cross-review skill
cross-review sends plans and code to a second model for review: Codex, Claude Code or Antigravity checks your agent's work.
- 167 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The reviewer runs with full system and network access; the installer edits agent configurations and project code goes to the second provider's account.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/cross-reviewDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The reviewer runs with full system and network access; the installer edits agent configurations and project code goes to the second provider's account.
This is third-party code. Review the repository files before installing.
What it does
cross-review installs six skills for Claude Code, Codex CLI and Antigravity. The agent you work with drives the task while a second model checks its plans, code and reasoning under a compiled three-layer Principal Reviewer system prompt. Codex CLI reviews by default, and `claude` or `agy` can be chosen instead. The skills: `loop-plan` writes a phased plan, `loop-implement` executes an approved plan with TDD and checkpoint reviews, `review-plan`, `review-impl` and `review-pr` check a plan, changes and someone else's pull request, and `review-think` runs a debate on a technical topic. Review progress shows on the local `cross-review dashboard`.
Who it is for. People who code with Claude Code or Codex and want a second pair of eyes on the plan and diff before committing.
Good fit when
- A large task is split into phases and the plan deserves a check before the first line of code.
- Before committing you want another model's findings on uncommitted changes.
- Someone else's pull request has to be reviewed for correctness and merge readiness.
- An agent debates a contested decision with the reviewer and the debate is recorded round by round.
Not a fit when
- Code cannot go to a second CLI, because it ends up in another provider's account.
- Only a quick look at a tiny edit is needed and a full review cycle would take longer than the edit.
- The environment has no Node.js 22 and no signed-in reviewer CLI.
Example request
Write a plan for refactoring the payments module, send it to Codex for review and show the findings round by round.Limitations
Node.js 22 or newer and at least one signed-in reviewer CLI (`codex`, `claude` or `agy`) are required. The reviewer runs with full access to run tests and network commands that check its findings; its prompts forbid changing project files, but the ban rests on an instruction. The installer writes to `~/.claude/skills`, `~/.agents/skills` and `~/.gemini/config` and registers an MCP server, and the `--auto` flag appends a block to `~/.claude/CLAUDE.md`. Installing removes the old codex-review pack; reviews still running under the old names cannot be resumed.
How to disable. Delete the `cross-review` skill directories from `~/.claude/skills` and `~/.agents/skills` and remove the MCP server with `claude mcp remove cross-review` and `codex mcp remove cross-review`. Remove the `--auto` block from `~/.claude/CLAUDE.md` by hand.
FAQ
How do I choose the reviewer?
Name it in the request (review this with Claude Code) or set `"reviewer"` in `~/.cross-review/config.json`. `codex`, `claude` and `agy` fit; Codex reviews by default.
How do I call the skills?
In Claude Code and Antigravity with `/name`, for example `/review-plan`, in Codex with `$name`.
Where do sessions and plans live?
In `<repo>/.cross-review/`, which ignores itself in git. Plans reach git only with the Plans in git setting on.
What does the --auto flag do?
It adds a rule to `CLAUDE.md` and `GEMINI.md` to run `review-plan` after a plan and `review-impl` after an implementation without asking. Enable it on purpose.
Which versions are needed?
Node.js 22 or newer and at least one signed-in CLI among `codex`, `claude` and `agy`.
Discussion
No questions or reviews yet.
Related
SkillsA skill for second-agent review of plans or code: Claude Code starts rounds, Codex reviews, and they exchange files
A Claude Code plugin: Claude does the task, then a stop hook starts an independent Codex review and Claude fixes the findings in the same session
The /improve skill: a strong model audits a repository and writes self-contained fix plans, and you hand the execution to a cheaper model
kyy-logs code-review-skill
code-review-skill
A code review skill with four independent probes, evidence-backed findings and a merge gate derived from JSON
A skill and CLI for code review: a team of reviewer agents inspects changes independently, debates findings and produces one report
Two skills: a pull request review where two models argue, and a helper that works through review comments round by round
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| cross-review (this card) | no | 167 | |
| Claude-Codex Review Skill | no | 5 | a Claude and Codex review skill; cross-review adds six skills, rounds with a findings ledger, a dashboard and a choice of reviewer among three CLIs |
| open-code-review | no | 49,561 | reviews code with one model; cross-review installs cross-checking in which a second model debates with the agent |
README in short
One `npx` command spreads the skills across three agents, installs a shared runner and registers an MCP server. Reviews run in rounds, findings collect in a ledger and a dashboard shows progress.