pr-review-fix-loop

The PR Review-Fix Loop is a controller-level agent skill designed for automated pull request maintenance. It functions by orchestrating specialized subagents in parallel to diagnose issues and sequentially to implement fixes, effectively reducing the manual overhead of code review cycles and CI maintenance. This skill is intended for software engineers and maintainers who want to streamline the iterative process of bringing branches to a deployable or mergeable state.

• Executes parallel review phases using code-reviewer models, CI status checks via gh pr checks, and comprehensive PR comment monitoring across inline, review, and issue threads.

• Supports multi-source issue aggregation, polling GitHub APIs for CodeQL alerts, linting results, and automated test outcomes to ensure no actionable feedback is missed.

• Manages an iterative loop (defaulting to 10 iterations) where a fixer subagent receives the aggregated issue list, pushes commits, and waits for CI triggers.

• Implements a safety-first controller pattern: the agent never performs file modifications or direct code edits itself, acting strictly as an orchestrator that delegates tasks to subagents with specific capabilities.

• Resolves target PRs via direct argument input, branch-head inspection, or interactive user prompts if a PR number is not initially provided.

• Best used when a PR is stuck in a failing CI state or requires resolution of multiple distinct code review comments.

• Requires appropriate authentication via GitHub CLI/Tokens to interact with the repository, pull requests, and commit histories.

• Operates on a 'fresh context' basis for every loop iteration, ensuring that subagents respond to the most recent state of the code and the latest CI/CD feedback.

• Designed for environments using standard GitHub workflows; integrates seamlessly with existing repository configurations.

• Constraints: The controller relies on the success of subagent reports; users should ensure that subagent prompts are correctly scoped for their project's tech stack and linting requirements.