adversarial-review Skill
Run a multi-model adversarial review loop over the current diff or requested scope. Use one Claude-family reviewer and one GPT-family reviewer in parallel, fix only high-confidence consensus issues, then rerun until no new consensus issues remain. Surface disputed findings to the user for a judgment call instead of aut Published by microsoft in XBOX-Godot-Sample.
What is adversarial-review Skill?
Run a multi-model adversarial review loop over the current diff or requested scope. Use one Claude-family reviewer and one GPT-family reviewer in parallel, fix only high-confidence consensus issues, then rerun until no new consensus issues remain. Surface disputed findings to the user for a judgment call instead of aut Published by microsoft in XBOX-Godot-Sample. This profile combines repository metadata with install, compatibility, and usage signals so developers can quickly decide whether it fits their agent workflow before opening the source repository.
Automated repository signals based on public metadata such as recency, license, installation evidence, and adoption. These are not a security audit or endorsement.
Key capabilities
- Includes SKILL.md support
- Reusable instructions support
- Testing
- Testing use cases
Technical details
- Install or run with Copy skill directory
When to use adversarial-review Skill
- Use it for testing.
Built with
Editorial notes
Source
- Creator: microsoft
- Repository: microsoft/XBOX-Godot-Sample
- Skill file: .github/skills/adversarial-review/SKILL.md
What it does
Run a multi-model adversarial review loop over the current diff or requested scope. Use one Claude-family reviewer and one GPT-family reviewer in parallel, fix only high-confidence consensus issues, then rerun until no new consensus issues remain. Surface disputed findings to the user for a judgment call instead of aut
Skill instructions
Adversarial Review Skill You are the repo-local adversarial-review skill. Your job is to pressure-test a change with two different model families, converge on the issues both families agree are real, fix the safe ones, and leave the disputed calls to the user. Core rules - Always use two model families in each review round: - one Claude-family model - one GPT-family model - Prefer the task tool with agenttype: "code-review" and explicit model overrides for the review passes. - Keep findings high-signal: bugs, contract drift, validation gaps, unsafe assumptions, resource lifetime issues, concurrency issues, and doc or spec drift. - Ignore style-only or cosmetic feedback. - Treat a finding as consensus only when both model families report materially the same issue. - Fix only consensus issues that are concrete, local, and unlikely to depend on product-direction choices. - Do not auto-fix disputed, speculative, or low-confidence findings. - After fixing consensus issues, rerun the same tw
Explore related resources
Frequently asked questions
What is adversarial-review?
adversarial-review is a open-source AI agent skill with Copy skill directory. Run a multi-model adversarial review loop over the current diff or requested scope.
Who is adversarial-review best for?
adversarial-review is best for reusing agent instructions, scripts, and references, testing workflows.
How do I install adversarial-review?
Install or run adversarial-review using Copy skill directory. Check adversarial-review for the latest setup command.
Is adversarial-review actively maintained?
adversarial-review may need a closer maintenance check before production use.
Auto-fetched from GitHub.
