Automated summary
This skill has the host agent call the opposite reviewer CLI (Codex when the host is Claude Code, and vice versa) to get an external assessment of a proposal or diff, then reconcile that assessment with its own inspection.
Strengths
- Forces an independent evaluation by a different model instead of a self-review that tends to agree with itself.
- Structures a host-written context note (intent, constraints, assumptions, points of uncertainty) so the external reviewer is not guessing.
- Focuses the request on reasonableness, regression risk, and alternatives rather than style nits.
- Produces auditable artifacts (prompt.md, assessment.md, reviewer.txt, run-summary.txt) written outside the repository.
Limitations
- Requires both the Claude Code and Codex CLIs to be installed and authenticated, since auto mode picks whichever is opposite the host.
- If the external reviewer times out or fails, you still have to sort through the logs and output files, and the assessment may be missing or partial.
- No automatic verdict: the host agent must arbitrate between the external opinion and its own direct inspection.
Use it when a design proposal or non-trivial diff needs an independent second opinion on soundness and risk before merging.
Skip it for style or formatting checks or one-line changes, where a round-trip to a second model adds no value.
Security analysis
CautionThe skill is an orchestration wrapper: it inspects git diffs and pipes them, plus a host-generated context note, into an external reviewer CLI. There is no destructive command, no credential exfiltration pattern, no obfuscated payload, and no attempt to disable safety controls; temp artifacts are explicitly kept out of the repository. The residual risk is inherent to the design rather than malicious: repo diffs and context are sent to an external model provider, and the actual work is done by a shell script whose contents are not included for inspection. This warrants 'caution' rather than 'safe' or 'risky'.
- •Delegates execution to an unshown helper script (scripts/run-cross-review.sh) that spawns external CLIs (codex/claude) and reads repo diffs
- •Invokes third-party AI CLIs which may transmit repository diff and host context to external services
- •Reads/creates files under a 'host-private system directory' whose exact path is not specified
Examples
Run a cross review on my staged changes: is this approach reasonable, what are the main risks, and is there a simpler alternative?交叉评审一下这个方案:我想把鉴权中间件换成基于 token 的方式,帮我让另一个 reviewer 评估风险点和有没有更强的替代方案。Use the opposite reviewer CLI to assess this PR diff for hidden regressions and missing constraints, then give me one reconciled final answer.name: cross-review description: Run a cross-review using the opposite CLI reviewer for proposal review and change assessment. Use when the user asks for cross review, 交叉 review, 交叉评审, second-opinion review, Codex-vs-Claude review, or wants the current host agent to ask the other reviewer whether a design or code change is reasonable, what risks it creates, and whether there is a simpler or stronger alternative.
Cross Code Review
Use this skill when the user wants the current host agent to call the other reviewer:
- if the current host is Codex, call Claude Code CLI
- if the current host is Claude Code, call Codex CLI
This skill is for:
- proposal review
- change assessment
- risk and regression analysis
- alternative-solution evaluation
It is not a generic code-style review skill.
Mandatory References
Before doing anything else, read these files:
These files are intentionally compact. They are guardrails, not the main payload.
Workflow
- Determine the review scope:
- user-specified files first
- otherwise inspect the current PR diff first
- otherwise inspect directly available repository diff
- otherwise ask what to review
- Prepare a short host context note that includes:
- what is changing or being proposed
- why this approach was chosen
- constraints, assumptions, and tradeoffs
- where you are uncertain
- Do this automatically from your current understanding of the task; do not ask the user to write it.
- Run the helper script to collect one external assessment from the opposite reviewer:
- auto mode:
scripts/run-cross-review.sh --repo <repo> --staged - working tree diff:
scripts/run-cross-review.sh --repo <repo> --working - one-click with auto-generated context note:
cat <<'EOF' | scripts/run-cross-review.sh --repo <repo> --staged --context-stdin --question 'Is this change reasonable? What are the main risks and is there a better alternative?' [host-generated context note] EOF - with context:
scripts/run-cross-review.sh --repo <repo> --staged --context-file /tmp/review-context.md - with question:
scripts/run-cross-review.sh --repo <repo> --staged --question 'Is this change reasonable? Is there a better alternative?' - with timeout override:
scripts/run-cross-review.sh --repo <repo> --working --timeout-seconds 120 - explicit Claude:
scripts/run-cross-review.sh --repo <repo> --staged --reviewer claude - explicit Codex:
scripts/run-cross-review.sh --repo <repo> --staged --reviewer codex
- auto mode:
- Read the generated files:
prompt.mdassessment.mdreviewer.txtrun-summary.txt- These files are written to a host-private system directory by default, not into the repository.
- The helper now prefers reviewing from the host context note and diff snapshot first. It should only inspect extra repository files when the diff alone is insufficient.
- Reconcile that external assessment with your own direct inspection using
references/integration.md. - Produce one final answer focused on reasonableness, risks, alternatives, and validation.
Rules
- In
--reviewer automode, if the current host is Codex, invoke Claude CLI. - In
--reviewer automode, if the current host is not Codex, invoke Codex CLI. - Prefer passing a host-written context note over relying on repository contents alone.
- Prefer
--context-stdinwhen the current agent is generating the note on the fly. - Prefer
--workingwhen the change is not staged yet. - Ask the external reviewer to evaluate whether the proposal or change is reasonable.
- Ask the external reviewer to identify hidden risks, missing constraints, and better alternatives.
- The helper has a timeout. If the external reviewer times out or fails, inspect
assessment.md,assessor-run.log, andrun-summary.txt, then continue with your own direct assessment. - Do not turn this into a generic code-style checklist.
- Do not copy the external assessment blindly; inspect the code and adjudicate.
- If the external CLI run fails, disclose it and continue with direct inspection.
- Keep temporary artifacts out of the repository by default.
Prerequisites
At least one external reviewer CLI must be installed and authenticated:
codexclaude
Which one is required depends on the selected reviewer target.
From the file to a first result
The download contains exactly the content displayed on this page. This copy may differ from the latest version in the source repository.
View the source and supporting filesThis download contains SKILL.md only. Any scripts, references, templates and dependencies are not included. Get the required files from the source, preserving their directory structure, before continuing.
1. Download the file
Download SKILL.mdFrom the root of a test project, save the file as .skill-downloads/cross-code-review-mrx3y0ri/SKILL.md, or use the command below. This folder is for inspecting the skill before installation.
Command for macOS/Linux or WSL: requires a POSIX shell and curl. On Windows without WSL, use the download button and create the folder manually.
mkdir -p '.skill-downloads/cross-code-review-mrx3y0ri' && curl --fail --show-error --location 'https://skills-guides.com/api/skills/cross-code-review-mrx3y0ri/download' --output '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md.part' && mv '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md.part' '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md'2. Check prerequisites
Read SKILL.md before using it. If it references scripts or external services, inspect them and prepare the required access in your test environment.
Original excerpt: Prerequisites
At least one external reviewer CLI must be installed and authenticated:
codexclaude
Which one is required depends on the selected reviewer target.
The platforms on this page are catalog indications. The paths below follow tool documentation; they do not prove this skill has been run in those tools. No minimum tool version is confirmed for this listing.
3. Load the skill
After inspection, copy the complete folder into your project using one of the commands below, then open a new tool session. The command stops if a folder with the same name already exists: compare the files before replacing it.
Claude Code
test -f '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md' && test -s '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md' && mkdir -p '.claude/skills' && test ! -e '.claude/skills/cross-review' && cp -R '.skill-downloads/cross-code-review-mrx3y0ri' '.claude/skills/cross-review'Codex
test -f '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md' && test -s '.skill-downloads/cross-code-review-mrx3y0ri/SKILL.md' && mkdir -p '.agents/skills' && test ! -e '.agents/skills/cross-review' && cp -R '.skill-downloads/cross-code-review-mrx3y0ri' '.agents/skills/cross-review'OpenSkills
OpenSkills must already be installed. The local install/read procedure is checked with OpenSkills 1.5.0, which requires Node.js ≥ 20.6.0. Catalog slugs are not resolved automatically.
Official OpenSkills installationopenskills install './.skill-downloads/cross-code-review-mrx3y0ri'openskills read 'cross-code-review-mrx3y0ri'Reading should display the installed file content and its base directory. It does not run the skill or install its dependencies.
For an agent that reads AGENTS.md, then sync the index. This command changes AGENTS.md: review the diff before using it.
openskills syncManual method or another tool: attach SKILL.md to a conversation, or open it in an agent that can read files. Include any required supporting resources. No universal folder is recognized by every tool.
4. Try a first task
Use a short, synthetic input: an API endpoint, a small document or an interface brief. In a new conversation, attach the file or give the agent its path, then adapt this request:
Read .skill-downloads/cross-code-review-mrx3y0ri/SKILL.md and apply its instructions to this example: [describe a test input]. Identify any missing files, dependencies or access. Produce the result requested by the skill without inventing data or running unreviewed scripts.Expected result: output matching the skill instructions, or a precise list of missing prerequisites. A read confirmation alone does not validate a first use. If an external resource is unavailable, that path remains untested.
React Hook Form Best Practices
Development
Master React Hook Form: avoid anti-patterns, manage subscriptions, and handle reset, dirty state, and controlled inputs correctly.
Emotion to Mantine Migration
Development
Migrate Emotion styled-components to Mantine components with style props and CSS modules. Use when converting .styled.tsx files or removing @emotion imports.
Scope Creep Detector
Development
Analyzes git diffs against a stated intent to detect scope creep, unrelated files, broad PRs, and oversized changes, then suggests keep, split, or justify actions.