Summary

Humanize tackles the challenge of achieving high-quality AI-generated code through an iterative development process featuring independent AI review. It targets developers aiming to build with confidence, leveraging continuous feedback loops for incremental code refinement and minimizing blind spots to meet acceptance criteria.

How It Works

The core is RLCR (Ralph-Loop with Codex Review), enhancing the ralph-loop concept with independent Codex code quality assessment. The process involves two phases: Implementation (Claude generates code, Codex reviews summaries) and Code Review (Codex checks code quality with severity markers). Issues identified feed back into implementation, creating a continuous refinement cycle. This "Iteration over Perfection" approach avoids single-shot expectations, and "Swarm Mode" allows parallel development with agent teams.

Quick Start & Requirements

• Installation: Via Humanize marketplace: humania marketplace add humania-org/humanize. For experimental features, use the dev branch: humania marketplace add humania-org/humanize#dev. Then, install the plugin: /plugin install humanize@humania.
• Prerequisites: Requires codex CLI. Refer to the Installation Guide for detailed prerequisites and alternative setup options.
• Quick Start:
  • Generate plan: bash /humanize:gen-plan --input draft.md --output docs/plan.md
  • Start RLCR loop: bash /humanize:start-rlcr-loop docs/plan.md
  • Monitor: source <path/to/humanize>/scripts/humanize.sh then humanize monitor rlcr.
• Documentation: Usage Guide, Installation Guide, Claude Code setup, Codex setup, Kimi CLI setup.

Highlighted Details

• "LLM IS AS GOOD AS YOU ARE" - aims to elevate AI code generation quality.
• "Iterative development with independent AI review" via Claude Code plugin.
• "Reinforcement Learning with Code Review" - iterative cycle of AI-generated code refined through external review feedback.
• "Swarm Mode" enables parallel development with agent teams.

Maintenance & Community

The provided README does not detail contributors, sponsorships, partnerships, or community channels (e.g., Discord/Slack).

Licensing & Compatibility

• License: MIT.
• Compatibility: MIT license generally permits commercial use and integration into closed-source projects.

Limitations & Caveats

• Requires codex CLI installation and configuration for the review process.
• The presence of a #dev branch for experimental features indicates potential instability or ongoing development unsuitable for production.