Extract and categorize GitHub PR reviewer comments into markdown training test cases.
npx skills add https://github.com/thomasttvo/claude-skills --skill capture-review
Please help me install this Agent Skill. Skill: capture-review Source: https://github.com/thomasttvo/claude-skills/tree/main/skills/capture-review Command: npx skills add https://github.com/thomasttvo/claude-skills --skill capture-review
Capture noisy or valuable reviewer insights from GitHub pull requests and transform them into consistent, structured training test cases so review agents improve over time and human signals are preserved.
Provide a GitHub PR URL and confirm the categorized comment list to save a structured training test case for that PR.
High-intent search queries and answers about installing and using this skill.
To save GitHub PR review comments as training data, fetch the pull request metadata and reviewer comments, filter for a specific reviewer's top-level messages, classify them into actionable signals versus context, and write the structured output to a local markdown test case file.
The best way to extract actionable signals from pull request review comments is to filter out reply threads and acceptance language, categorize the remaining top-level feedback into signals like nits or issues, and mark outdated threads using original commit SHAs.
Yes, you can generate structured test cases from GitHub pull request feedback without modifying the repository because the process only fetches PR metadata and review comments to write or append a local markdown training file without pushing changes back to GitHub.
To handle outdated review threads when extracting training data from pull requests, detect the threads and capture their original commit SHAs so the generated test cases accurately reference the specific code state the reviewer commented on.
Yes, this PR comment extraction method supports appending new feedback to existing markdown training files by mapping user selections back to comment IDs or ranges and writing or appending the categorized test case into a local training directory.
A key limitation when categorizing GitHub review comments into severity labels for model training is that the extraction only captures top-level comments from a specific target reviewer, intentionally filtering out reply threads and general acceptance language to focus on distinct signals.
samcdavidPublish a GitHub PR review with inline comments and thread replies.
xobotyiFormat code review findings into a standardized Markdown structure.
nicolehaugenGenerate pull request review comment replies linked to commit SHAs.
furkankoykiranGenerate GitHub comments referencing conversation context and thread tone.
