zevorn/dsh-humanize
30Last commit Aug 13, 2026
dsh-humanize DSH plugin
Humanize is a plugin for Claude Code that implements the RLCR (Ralph-Loop with Codex Review) workflow. It enables iterative development where Claude implements code and Codex independently reviews it, with feedback loops until all acceptance criteria are met.
How to install the dsh-humanize DSH plugin
dsh plugin --profile web add github:dsh-external/dsh-humanize#<commit-or-tag>This source command needs manual review. Copying does not run it.
dsh-humanize DSH plugin data source
dsh-humanize DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-humanize DSH plugin can do
- Generate idea drafts from loose thoughts and explore multiple directions as parallel prototypes
- Generate structured plans with capability maps and optional coaching quizzes
- Run RLCR loops that automatically iterate between implementation and code review
- Consult Gemini for deep web research (requires Gemini CLI)
- Monitor progress in a separate terminal with specific commands (rlcr, skill, codex, gemini)
Where the dsh-humanize DSH plugin fits
- Iteratively develop a new feature with continuous code review to catch issues early
- Refactor existing code while ensuring quality through independent review
- Explore multiple design directions in parallel via isolated git worktrees
- Generate and refine plans before writing code to ensure up-front design alignment
Who the dsh-humanize DSH plugin is for
- Developers using Claude Code who want to improve code quality through AI review
- Teams adopting an iterative development workflow with built-in review gates
dsh-humanize DSH plugin limitations
- Requires codex CLI from OpenAI for the review step
- Gemini CLI is optional but required for the ask-gemini skill
- Plugin is still in early development (version 1.18.0)
- Monitoring requires a separate terminal outside Claude Code
dsh-humanize DSH plugin: from the repository README
Quoted from the zevorn/dsh-humanize README, the upstream source of the dsh-humanize DSH plugin. Copyright remains with the original authors.
**Current Version: 1.18.0** > Derived from the [GAAC (GitHub-as-a-Context)](https://github.com/SihaoLiu/gaac) project. A Claude Code plugin that provides iterative development with independent AI review. Build with confidence through continuous feedback loops. ## What is RLCR? **RLCR** stands for **Ralph-Loop with Codex Review**, inspired by the official ralph-loop plugin and enhanced with independent Codex review. The name also reads as **Reinforcement Learning with Code Review** -- reflecting the iterative cycle where AI-generated code is continuously refined through external review feedback. ## Core Concepts - **Iteration over Perfection** -- Instead of expecting perfect output in one shot, Humanize leverages continuous feedback loops where issues are caught early and refined incrementally. - **One Build + One Review** -- Claude implements, Codex independently reviews. No blind spots. - **Ralph Loop with Swarm Mode** -- Iterative refinement continues until all acceptance criteria are met. Optionally parallelize with Agent Teams. - **Capability Anchors** -- Generated plans include a feature/capability map, and RLCR rounds keep Claude and Codex anchored to the relevant capab
Read the full READMEThe repository declares no license. Check with the authors before using it.
dsh-humanize DSH plugin questions
How do I install Humanize for DSH?
Run the command: `dsh plugin --profile web add github:dsh-external/dsh-humanize#<commit-or-tag>`. If pnpm blocks the build, add the exact package key to `$DSH_HOME/profiles/web/pnpm-workspace.yaml` under `allowBuilds`, then rerun the command. Configure the builder model (`deepseek-v4-flash-max`) in DSH model settings.
What dependencies are required for Humanize?
Humanize requires the codex CLI from OpenAI for the review step. Gemini CLI is optional but needed for the ask-gemini skill. For DSH installation, you need pnpm and the DSH web profile set up.
How do I monitor the RLCR loop progress?
Open a separate terminal (not inside Claude Code) and source the humanize.sh script. Then use commands like `humanize monitor rlcr` to monitor the RLCR loop, `humanize monitor skill` for all skill invocations, or `humanize monitor codex` for codex-specific monitoring.
What is the RLCR workflow?
RLCR stands for Ralph-Loop with Codex Review. It's an iterative cycle where Claude implements code, Codex independently reviews it, and feedback feeds back into implementation until all acceptance criteria are met. The process includes two phases: Implementation and Code Review.
Can I use Humanize without Claude Code?
Humanize is specifically designed as a Claude Code plugin. It requires Claude Code to run the implementation agent. However, it can be installed as a DSH profile bundle, which runs the RLCR loop inside a DeepSeek Harness session with a different builder agent.