Skillget

harness-engineering-skills

skill
Security Audit
Fail
Health Warn
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Fail
  • rm -rf — Recursive force deletion command in plugins/harness-engineering-skills/skills/review-loop/scripts/peer-invoke.sh
Permissions Pass
  • Permissions — No dangerous permissions requested
Purpose
This tool provides cross-LLM iterative code review and multi-agent orchestration skills for Claude Code. It spawns external peers like Codex or Gemini to independently review code changes, automatically implementing fixes until a consensus is reached.

Security Assessment
The tool inherently executes shell commands to spawn external CLI processes (Codex, Gemini) and manages file structures via git. A notable security flag is the presence of a recursive force deletion command (`rm -rf`) inside its shell scripts, which could be dangerous if path variables are unexpectedly modified. It requires external network access to interact with other AI models, but no hardcoded secrets were detected, and it does not request explicitly dangerous broad permissions. Because of the `rm -rf` execution and multi-process spawning, the overall risk is rated as Medium.

Quality Assessment
The repository is actively maintained, with its most recent push occurring today. It is backed by the standard Apache-2.0 license, and the documentation is thorough, offering clear installation and usage guides. However, community trust and visibility are currently very low. With only 6 GitHub stars, the tool appears to be a niche, early-stage project that has not yet been broadly vetted by the developer community.

Verdict
Use with caution: while the repository is active and legally clear, the inherent shell execution risks and low community adoption warrant careful review before integrating into sensitive workflows.
SUMMARY

Cross-LLM iterative code review plugin for Claude Code. Spawns Codex or Gemini as peer reviewer, loops until consensus.

README.md

English | 中文

Harness Engineering Skills

Stometa's public curated Claude Code skillset — a small, opinionated set of skills we use ourselves, published periodically.

License
Claude Code
Peer: Codex
Peer: Gemini

Why this repo

This is the public companion to Stometa's private stometa-skillset. We dogfood a larger internal skillset day-to-day; selected skills are extracted, polished, and published here in batches. The goal is to share the workflows that actually hold up under real engineering work — not a pile of prototypes.

The first batch ships two skills: review-loop (already proven in daily use) and harness (multi-agent orchestration for larger tasks). Both are installable as a single Claude Code plugin.

Skills

review-loop

Cross-LLM iterative code review. Spawns a peer reviewer (Codex CLI or Gemini CLI) to independently review your changes. Claude evaluates the peer's findings, implements accepted fixes, and re-submits until both sides agree on the final code state. The human doesn't need to participate — watch progress via .review-loop/<session>/summary.md.

harness

Cybernetics-based multi-agent orchestration for complex tasks. Coordinates a Planner → Generator → Evaluator → Retro pipeline with fresh sub-agents per checkpoint (drift prevention) and persistent retro learning across tasks. Recommended flow: Claude Code plans the spec (Session 1), Codex executes autonomously (Session 2), and review-loop (Codex or Gemini CLI as peer) provides the cross-model quality gate before PR.

Install

claude plugin marketplace add https://github.com/stone16/harness-engineering-skills
claude plugin install harness-engineering-skills@stometa

Verify:

claude plugin list | grep harness-engineering-skills

Prerequisites

  • Required: git, python3, Claude Code with the superpowers plugin installed.
  • Peer reviewer (one of): codex CLI or gemini CLI — only needed if you use review-loop or harness's cross-model review.
  • Optional: gh CLI for PR-scoped review detection.

Usage

review-loop — inside a Claude Code session, once the plugin is installed:

/review-loop

Variants: review loop with gemini, review loop, max 3 rounds, review loop for PR 42, review loop for commit abc123.

harness — start a new orchestrated task:

harness plan <task-id>

Then follow the planning dialogue; harness will drive checkpoints through Generator → Evaluator with a cross-model review gate before producing a PR.

License

Apache-2.0 — see LICENSE.

Origin and related

This repo is the public publication surface for a subset of Stometa's private stometa-skillset. Future batches will add more skills as they stabilize. Issues and pull requests are welcome on the GitHub tracker.

Reviews (0)

No results found