gitlab-address-comments_skill

Overview

This skill helps find the open GitLab Merge Request (MR) for the current branch and address review or issue comments using the glab CLI. It automates fetching unresolved discussions, summarizing and triaging them, and applying focused fixes in-repo. The workflow surfaces the most relevant threads, requests a single batch selection, and then runs changes and checks before reporting results.

How this skill works

The skill first verifies glab authentication and locates the MR for the current branch. It uses a local helper script that calls glab api to fetch MR discussions, filters out bot/system-only threads, and writes unresolved discussions to a temporary JSON. The assistant summarizes threads, asks the user to pick a batch, shows full context for selected items, applies targeted fixes, runs available tests/checks, and reports which discussions were addressed.

When to use it

• You have an open GitLab MR on the current git branch and want to address review comments efficiently.
• You need to triage many unresolved discussions and pick a focused subset to fix in one run.
• You want code-context (±3 lines) when addressing comments to avoid context-switching.
• You want to run local tests or checks after applied fixes and report back changes.
• You need to convert reviewer threads into concrete, reproducible code edits.

Best practices

• Ensure glab auth is valid (glab auth status) and glab >= v1.80.4 before running.
• Run the fetch script with --open-comments to focus only on unresolved threads.
• Accept a single batch selection (e.g., 1,3,5-7 or top N) instead of answering per comment.
• Group summaries by file path and present one-line summaries to reduce noise.
• Run the most relevant test suites or linters after edits and report test outcomes.

Example use cases

• A contributor wants to quickly resolve the top three unresolved review comments on a feature branch.
• A maintainer needs a compact list of unresolved MR discussions grouped by file to assign work.
• A developer wants precise code context for each selected thread before making changes.
• An automated assistant fetches unresolved threads, applies fixes, runs tests, and posts a status update.