self-review-loop

Coordinate iterative PR reviews with fresh sub-agents via GitHub CLI.

Updated Mar 22, 2026

Search Tags:#automation #orchestration #pr #code-review #pull-request #ai-assisted #self-review

Ready to run

npx skills add https://github.com/pamelia/opencode-plugins --skill self-review-loop

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: self-review-loop
Source: https://github.com/pamelia/opencode-plugins/tree/main/skills/self-review-loop
Command: npx skills add https://github.com/pamelia/opencode-plugins --skill self-review-loop

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Iterative PR self-review orchestration that launches a fresh, context-free sub-agent per turn to perform a code review against the PR diff, ensuring unbiased evaluation without anchoring to prior feedback.

Core Features & Use Cases

Fresh, context-free review agents per turn to independently evaluate PR changes
Aggregates and applies feedback across turns, stopping when only non-actionable feedback remains
Oscillation detection and termination to prevent undo/redo cycles; maintains changelog and state per PR
Auto-discovers available code-review skills and coordinates multi-turn PR review workflows

Quick Start

Instantiate the self-review-loop skill against a PR to begin iterative reviews.

Frequently Asked Questions about self-review-loop

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I automate iterative code review for a pull request?▼

Automated PR review is orchestrated by launching fresh, context-free sub-agents to evaluate the diff across multiple turns. It aggregates feedback, applies changes, and stops when only non-actionable feedback remains or after five turns.

Why does my AI code review keep repeating the same feedback?▼

AI code review repetition is prevented by oscillation detection and termination logic. The orchestration loop identifies undo/redo cycles in PR feedback and halts the process to avoid redundant iterations.

Do I need a code-review skill installed to automate PR feedback?▼

Yes, an installed code-review capability is required to automate PR feedback. The orchestration loop auto-discovers available code-review skills and coordinates them to evaluate pull request diffs.

Can I use GitHub CLI to access pull requests for automated review?▼

GitHub CLI is used to access pull requests for automated review. The orchestration loop relies on it to retrieve PR diffs, manage loop state, and apply feedback changes across multiple review turns.

What is the maximum number of turns for an iterative PR review loop?▼

The maximum number of turns for an iterative PR review loop is five. The process terminates automatically after five turns, when only non-actionable feedback remains, or if oscillation is detected.

How do I prevent anchoring bias in automated PR code reviews?▼

Anchoring bias in automated PR code reviews is prevented by launching fresh, context-free sub-agents per turn. Each sub-agent independently evaluates the PR diff without prior feedback context, ensuring unbiased evaluation.