Repository navigation
feat: restructure review prompt with caveman pipeline approach - #19
Merged
Merged
Conversation
Separates internal reasoning (terse caveman-full) from output style (caveman-light). Reasoning mode compresses how Claude thinks through the diff without impacting comment readability. Output switches from full caveman to caveman-light — no filler or hedging, but full sentences and clear explanations preserved. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
Removes verbose prose from the template so projects inherit compact guidelines by default. No caveman-compress dependency needed — small files stay small naturally. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
anuragk16
marked this pull request as ready for review
April 15, 2026 14:14
This comment was marked as resolved.
This comment was marked as resolved.
HTML comment was invisible in rendered Markdown. Converted back to a blockquote so teams copying the template can see the multi-file guidance. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
anuragk16
added a commit
that referenced
this pull request
Apr 17, 2026
Reverts changes from PR #18 and PR #19 that applied caveman-style formatting to the code review workflow prompt and guidelines template. Co-Authored-By: Claude Sonnet 4.6 <[email protected]>
2 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
What changed
claude-code-review.yml— prompt restructured into 3 layers:bug:/risk:/nit:/q:per finding, format<file>:L<line>: <severity>: <problem>. <fix>.automated-code-review/templates/review-guidelines.md— rewritten terse:Why caveman-light over full caveman
Full caveman output (fragments, dropped articles) hurt PR comment readability with minimal extra token savings. Caveman-light keeps full sentences but removes all fluff — better tradeoff for human-facing output.
Token impact
Thinking tokens (billed same as output) are not controllable via prompt — model decides thinking depth based on complexity.
No new dependencies
Projects do not need to install caveman plugin or run any tool. All optimization is baked into the shared workflow prompt and the guidelines template.
Test plan
ready for reviewlabel on a test PR<file>:L<line>: [bug|risk|nit|q]:format🤖 Generated with Claude Code