evalstate/fast-agent

Overview

This skill extracts and analyzes editorial feedback from GitHub pull request review comments to help writers learn from changes. It produces structured before/after comparisons, catalogs explicit suggestions, and synthesizes style lessons to guide future writing. Use it to turn PR review noise into actionable writing improvements.

How this skill works

The tool fetches PR review comments and suggestion blocks, maps them to file versions across the PR, and extracts explicit suggestion pairs plus plain-text reviewer feedback. When requested, it also pulls FIRST and FINAL file contents and performs paragraph-by-paragraph comparisons to surface what changed and why. The output groups mechanical fixes, mapped feedback, and synthesized style patterns for easy consumption.

When to use it

• You want a concise summary of editorial feedback from a GitHub PR.
• You need before/after comparisons for text files changed in a PR.
• You want to catalog recurring style or grammar issues flagged by reviewers.
• You are preparing a style guide or onboarding notes from past PRs.
• You need lightweight outputs for LLM prompting or deeper diffs for manual review.

Best practices

• Run with --diff when you need first-to-final comparisons; add --max-file-chars to limit block size.
• Use --no-files if you only need suggestion summaries to keep results compact.
• Preserve reviewer permalinks to maintain context for follow-up questions.
• Group mechanical fixes separately from tone/structural feedback for clearer lessons.
• Treat the synthesized patterns as learning suggestions, not absolute rules.

Example use cases

• Generate a table of recurring grammar and capitalization fixes across PRs to inform a style checklist.
• Produce paragraph-level change logs to teach authors how reviewer requests map to concrete edits.
• Create a concise bundle of reviewer suggestions for an editor or content owner to act on.
• Feed capped FIRST/FINAL snippets into an LLM to generate higher-level writing lessons or templates.
• Trace how a sentence evolved from draft to final to demonstrate effective rewrites in onboarding.