Skip to content

shinezyy/humanize

 
 

Repository files navigation

Humanize

Current Version: 1.10.4

Derived from the GAAC (GitHub-as-a-Context) project.

What is RLCR?

RLCR stands for Ralph-Loop with Codex Review. It was inspired by the official ralph-loop plugin, enhanced with a series of optimizations and independent Codex review capabilities.

The name can also be interpreted as Reinforcement Learning with Code Review - reflecting the iterative improvement cycle where AI-generated code is continuously refined through external review feedback.

A Claude Code plugin that provides iterative development with Codex review. Humanize creates a feedback loop where Claude implements your plan while Codex independently reviews the work, ensuring quality through continuous refinement.

Core Philosophy

Iteration over Perfection: Instead of expecting perfect output in one shot, Humanize leverages an iterative feedback loop where:

  • Claude implements your plan
  • Codex independently reviews progress
  • Issues are caught and addressed early
  • Work continues until all acceptance criteria are met

This approach provides:

  • Independent review preventing blind spots
  • Goal tracking to prevent drift
  • Quality assurance through iteration
  • Complete audit trail of development progress

Installation

Option 1: Install from Git Marketplace (Recommended)

Start Claude Code and run the following commands:

# Add the marketplace
/plugin marketplace add git@github.com:humania-org/humanize.git

# Install the plugin
/plugin install humanize@humania

Option 2: Local Development / Testing

If you have the plugin cloned locally:

# Start Claude Code with the plugin directory
claude --plugin-dir /path/to/humanize

Prerequisites

  • codex - OpenAI Codex CLI (for review). Check with codex --version.

Environment Variables

Humanize supports the following environment variables for advanced configuration:

HUMANIZE_CODEX_BYPASS_SANDBOX

WARNING: This is a dangerous option that disables security protections. Use only if you understand the implications.

  • Purpose: Controls whether Codex runs with sandbox protection
  • Default: Not set (uses --full-auto with sandbox protection)
  • Values:
    • true or 1: Bypasses Codex sandbox and approvals (uses --dangerously-bypass-approvals-and-sandbox)
    • Any other value or unset: Uses safe mode with sandbox

When to use this:

  • Linux servers without landlock kernel support (where Codex sandbox fails)
  • Automated CI/CD pipelines in trusted environments
  • Development environments where you have full control

When NOT to use this:

  • Public or shared development servers
  • When reviewing untrusted code or pull requests
  • Production systems
  • Any environment where unauthorized system access could cause damage

Security implications:

  • Codex will have unrestricted access to your filesystem
  • Codex can execute arbitrary commands without approval prompts
  • Review all code changes carefully when using this mode

Usage example:

# Export before starting Claude Code
export HUMANIZE_CODEX_BYPASS_SANDBOX=true

# Or set for a single session
HUMANIZE_CODEX_BYPASS_SANDBOX=true claude --plugin-dir /path/to/humanize

Usage

How It Works

flowchart LR
    Plan["Your Plan<br/>(plan.md)"] --> Claude["Claude Implements<br/>& Summarizes"]
    Claude --> Codex["Codex Reviews<br/>Summary"]
    Codex -->|Feedback Loop| Claude
    Codex -->|COMPLETE| Review["Code Review<br/>(codex review)"]
    Review -->|Issues Found| Claude
    Review -->|No Issues| Done((Done))
Loading

The loop has two phases:

  1. Implementation Phase: Claude works, Codex reviews summaries until COMPLETE
  2. Review Phase: codex review --base <branch> checks code quality with [P0-9] severity markers

Quick Start

  1. Create a plan file or just write down your thoughts in <name/you/like/for/draft>.md and use /humanize:gen-plan
    /humanize:gen-plan --input <name/you/like/for/draft.md> --output <docs/my-feature-plan.md>
  2. Run the loop:
    /humanize:start-rlcr-loop <docs/my-feature-plan.md>
  3. Monitor progress in .humanize/rlcr/<timestamp>/ or you can use the monitor script:
    source ~/.claude/plugins/cache/humania/humanize/<LATEST.VERSION>/scripts/humanize.sh // Add this to your .bashrc or .zshrc
    humanize monitor [rlcr|pr] // Launch this from where you start claude to monitor RLCR loop or PR loop
  4. Cancel if needed: /humanize:cancel-rlcr-loop

Commands

Command Purpose
/start-rlcr-loop <plan.md> Start iterative development with Codex review
/cancel-rlcr-loop Cancel active loop
/gen-plan --input <draft.md> --output <plan.md> Generate structured plan from draft
/start-pr-loop --claude|--codex Start PR review loop with bot monitoring
/cancel-pr-loop Cancel active PR loop
/ask-codex [question] One-shot consultation with Codex

Command Options

start-rlcr-loop

/humanize:start-rlcr-loop [path/to/plan.md | --plan-file path/to/plan.md] [OPTIONS]

OPTIONS:
  --plan-file <path>     Explicit plan file path (alternative to positional arg)
  --max <N>              Maximum iterations before auto-stop (default: 42)
  --codex-model <MODEL:EFFORT>
                         Codex model and reasoning effort (default: gpt-5.3-codex:xhigh)
  --codex-timeout <SECONDS>
                         Timeout for each Codex review in seconds (default: 5400)
  --track-plan-file      Indicate plan file should be tracked in git (must be clean)
  --push-every-round     Require git push after each round (default: commits stay local)
  --base-branch <BRANCH> Base branch for code review phase (default: auto-detect)
                         Auto-detection priority: remote default > main > master
  --full-review-round <N>
                         Interval for Full Alignment Check rounds (default: 5, min: 2)
  --skip-impl            Skip implementation phase, go directly to code review
                         Plan file is optional when using this flag
  -h, --help             Show help message

gen-plan

/humanize:gen-plan --input <path/to/draft.md> --output <path/to/plan.md>

OPTIONS:
  --input   Path to the input draft file (required)
  --output  Path to the output plan file (required)

The gen-plan command transforms rough draft documents into structured implementation plans.

Workflow:
1. Validates input/output paths
2. Checks if draft is relevant to the repository
3. Analyzes draft for clarity, consistency, completeness, and functionality
4. Engages user to resolve any issues found
5. Generates a structured plan.md with AC-X acceptance criteria

start-pr-loop

/humanize:start-pr-loop --claude|--codex [OPTIONS]

BOT FLAGS (at least one required):
  --claude   Monitor reviews from claude[bot] (trigger with @claude)
  --codex    Monitor reviews from chatgpt-codex-connector[bot] (trigger with @codex)

OPTIONS:
  --max <N>              Maximum iterations before auto-stop (default: 42)
  --codex-model <MODEL:EFFORT>
                         Codex model and reasoning effort (default: gpt-5.2-codex:medium)
  --codex-timeout <SECONDS>
                         Timeout for each Codex review in seconds (default: 900)
  -h, --help             Show help message

ask-codex

/humanize:ask-codex [OPTIONS] <question or task>

OPTIONS:
  --codex-model <MODEL:EFFORT>
                         Codex model and reasoning effort (default: gpt-5.3-codex:xhigh)
  --codex-timeout <SECONDS>
                         Timeout for the Codex query in seconds (default: 3600)
  -h, --help             Show help message

The ask-codex skill sends a one-shot question or task to Codex and returns the response inline. Unlike the RLCR loop, this is a single consultation without iteration -- useful for getting a second opinion, reviewing a design, or asking domain-specific questions.

Responses are saved to .humanize/skill/<timestamp>/ with input.md, output.md, and metadata.md for reference.

The PR loop automates the process of handling GitHub PR reviews from remote bots:

  1. Detects the PR associated with the current branch
  2. Fetches review comments from the specified bot(s)
  3. Claude analyzes and fixes issues identified by the bot(s)
  4. Pushes changes and triggers re-review by commenting @bot
  5. Stop Hook polls for new bot reviews (every 30s, 15min timeout per bot)
  6. Local Codex validates if remote concerns are approved or have issues
  7. Loop continues until all bots approve or max iterations reached

Prerequisites:

  • GitHub CLI (gh) must be installed and authenticated
  • Codex CLI must be installed
  • Current branch must have an associated open PR

Monitoring:

humanize monitor pr

License

MIT

Credits

About

LLM is as good as you are

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages

  • Shell 99.5%
  • Python 0.5%