PluginBench
Skill
Review
Audit score 70

caveman-review

juliusbrussee/caveman

Compressed code review: one line per finding with location, problem, and fix.

SourceView on skills.sh

What is caveman-review?

A code review skill that enforces terse, actionable feedback. One line per finding with exact location, concrete problem statement, and specific fix—no hedging or throat-clearing. Use it to review PRs, diffs, or code snippets with maximum signal and minimum noise.

  • Format findings as `L<line>: <problem>. <fix>.` with optional severity prefix (🔴 bug, 🟡 risk, 🔵 nit, ❓ q)
  • Strip filler language: no "I noticed", "you might consider", or restating what the code does
  • Include exact line numbers, symbol names in backticks, and concrete fixes with rationale when non-obvious
  • Expand to full paragraphs only for security findings, architectural disagreements, or onboarding contexts
  • Output comments ready to paste directly into PR review tools

How to install caveman-review

npx skills add https://github.com/juliusbrussee/caveman --skill caveman-review
Claude Code
Cursor
Windsurf
Cline

How to use caveman-review

  1. 1.Invoke with `/caveman-review`, "review this PR", or "review the diff"
  2. 2.Paste the code or diff you want reviewed
  3. 3.Read the one-line findings with location, problem, and fix
  4. 4.For security or architectural issues, expect a full paragraph instead of a one-liner
  5. 5.Use "stop caveman-review" or "normal mode" to revert to verbose review style

Use cases

Good for
  • Review pull requests with rapid, focused feedback on bugs and risks
  • Audit diffs across multiple files with consistent severity labeling
  • Provide code review in high-velocity teams where verbose comments slow iteration
  • Identify fragile patterns (race conditions, missing null checks, swallowed errors) in one-liner format
  • Train junior developers on what actionable feedback looks like
Who it's for
  • Code reviewers in fast-moving teams
  • Engineering leads standardizing review culture
  • Developers using Claude Code or Cursor for PR feedback
  • Teams adopting terse communication norms

caveman-review FAQ

When should I use full paragraphs instead of one-liners?

For security findings (CVE-class bugs need references), architectural disagreements (need rationale), and onboarding contexts where the author is new and needs the "why". Resume terse mode for the rest of the review.

What do the severity prefixes mean?

🔴 bug = broken behavior causing incidents; 🟡 risk = works but fragile (race, null check, swallowed error); 🔵 nit = style/naming/micro-optimization (author can ignore); ❓ q = genuine question, not a suggestion.

Can I use this to write the code fix?

No. Caveman-review outputs comments only—it does not write code, approve/request changes, or run linters. You paste the output directly into the PR.

What if I'm unsure about a finding?

Use the `❓ q:` prefix to frame it as a genuine question rather than a suggestion. Avoid hedging language like "perhaps" or "maybe".

How do I revert to normal review style?

Say "stop caveman-review" or "normal mode" to exit terse mode and resume verbose review comments.

Full instructions (SKILL.md)

Source of truth, from juliusbrussee/caveman.


name: caveman-review description: > Compressed code review - one line per finding with location, problem and fix. Use for /caveman-review, "review this PR", or "review the diff".

Write code review comments terse and actionable. One line per finding. Location, problem, fix. No throat-clearing.

Rules

Format: L<line>: <problem>. <fix>. — or <file>:L<line>: ... when reviewing multi-file diffs.

Severity prefix (optional, when mixed):

  • 🔴 bug: — broken behavior, will cause incident
  • 🟡 risk: — works but fragile (race, missing null check, swallowed error)
  • 🔵 nit: — style, naming, micro-optim. Author can ignore
  • ❓ q: — genuine question, not a suggestion

Drop:

  • "I noticed that...", "It seems like...", "You might want to consider..."
  • "This is just a suggestion but..." — use nit: instead
  • "Great work!", "Looks good overall but..." — say it once at the top, not per comment
  • Restating what the line does — the reviewer can read the diff
  • Hedging ("perhaps", "maybe", "I think") — if unsure use q:

Keep:

  • Exact line numbers
  • Exact symbol/function/variable names in backticks
  • Concrete fix, not "consider refactoring this"
  • The why if the fix isn't obvious from the problem statement

Examples

❌ "I noticed that on line 42 you're not checking if the user object is null before accessing the email property. This could potentially cause a crash if the user is not found in the database. You might want to add a null check here."

✅ L42: 🔴 bug: user can be null after .find(). Add guard before .email.

❌ "It looks like this function is doing a lot of things and might benefit from being broken up into smaller functions for readability."

✅ L88-140: 🔵 nit: 50-line fn does 4 things. Extract validate/normalize/persist.

❌ "Have you considered what happens if the API returns a 429? I think we should probably handle that case."

✅ L23: 🟡 risk: no retry on 429. Wrap in withBackoff(3).

Auto-Clarity

Drop terse mode for: security findings (CVE-class bugs need full explanation + reference), architectural disagreements (need rationale, not just a one-liner), and onboarding contexts where the author is new and needs the "why". In those cases write a normal paragraph, then resume terse for the rest.

Boundaries

Reviews only — does not write the code fix, does not approve/request-changes, does not run linters. Output the comment(s) ready to paste into the PR. "stop caveman-review" or "normal mode": revert to verbose review style.