Automate independent reviews across Claude, Gemini, Codex, OpenCode, and Kimi.
npx skills add https://github.com/fkguo/nullius --skill review-swarm-fkguo
Please help me install this Agent Skill. Skill: review-swarm Source: https://github.com/fkguo/nullius/tree/main/skills/review-swarm Command: npx skills add https://github.com/fkguo/nullius --skill review-swarm-fkguo
This Skill streamlines the process of obtaining independent, multi-backend reviews for various types of content, ensuring accuracy and reliability.
To start a multi-backend review swarm, run the following command:
python3 scripts/bin/run_multi_task.py \
--out-dir /tmp/cross_family_review \
--system /path/to/reviewer_system.md \
--prompt /path/to/packet.md \
--models codex/default,gemini/default,zhipuai-coding-plan/glm-5.2 \
--check-review-contract
High-intent search queries and answers about installing and using this skill.
You can automate independent code reviews across multiple AI backends like Claude, Gemini, Codex, OpenCode, and Kimi by running a review swarm script that dispatches prompts to each model and records their distinct outputs.
A review contract check validates output format compliance during a multi-backend review swarm, ensuring that each AI backend returns results that meet predefined structural rules before recording them.
To run a cross-family review, execute the multi-task Python script with your system prompt, packet file, and a comma-separated list of target models, enabling the review contract flag to validate outputs.
Yes, you need backend-specific runner skills such as the claude-cli-runner, gemini-cli-runner, and codex-cli-runner available on your PATH to dispatch and execute the independent review tasks.
When an AI backend fails or returns invalid output during an automated review swarm, the system applies a fallback policy to handle the failure gracefully and ensure the overall review process continues.
Yes, you can gate the multi-backend review process on convergence by applying an optional Jaccard similarity check to measure the overlap and consistency of outputs from the different AI backends.
Solitude01Search, install, update, and publish agent skills via the clawhub CLI.
blueman82Diagnose deliberation convergence failures and output actionable remediation reports.
blocknaviCreate self-contained Convex components with isolated schemas and exports.
GuicedEEValidate Terraform configurations for syntax, formatting, security, and compliance.