Caveman review
Ultra-compressed code review comments. Cuts noise from PR feedback while preserving the actionable signal. Each comment is one line: location, problem, fix. Use when user says "review this PR", "code review", "review the diff", "/review", or invokes /caveman-review. Auto-triggers when reviewing pull requests.From its SKILL.md
npx -y skills add nota-america/forgecat-agent-profiles --skill caveman-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
2.7 KB, 593 tokens by cl100k_base, as published. Nobody here has run it
Write code review comments terse and actionable. One line per finding. Location, problem, fix. No throat-clearing.
Rules
Format: L<line>: <problem>. <fix>. β or <file>:L<line>: ... when reviewing multi-file diffs.
Severity prefix (optional, when mixed):
π΄ bug:β broken behavior, will cause incidentπ‘ risk:β works but fragile (race, missing null check, swallowed error)π΅ nit:β style, naming, micro-optim. Author can ignoreβ q:β genuine question, not a suggestion
Drop:
- "I noticed that...", "It seems like...", "You might want to consider..."
- "This is just a suggestion but..." β use
nit:instead - "Great work!", "Looks good overall but..." β say it once at the top, not per comment
- Restating what the line does β the reviewer can read the diff
- Hedging ("perhaps", "maybe", "I think") β 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
β "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."
β
L42: π΄ bug: user can be null after .find(). Add guard before .email.
β "It looks like this function is doing a lot of things and might benefit from being broken up into smaller functions for readability."
β
L88-140: π΅ nit: 50-line fn does 4 things. Extract validate/normalize/persist.
β "Have you considered what happens if the API returns a 429? I think we should probably handle that case."
β
L23: π‘ 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 β 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.
What ships with it: 1 file
1.2 KB alongside SKILL.md
- README.md1.2 KB
Gives 0 of the 12 instructions most code review skills give in 593 tokens
Counted across 668 of the 814 authors here whose files we hold, read 2026-09-06
- Provide technical reasoning when pushing backin 84 of 668, across 70 files
- Fix critical issues immediatelyin 77 of 668, across 60 files
- Dispatch a code reviewer subagentin 76 of 668, across 59 files
- Fix important issues before proceedingin 73 of 668, across 56 files
- Ask for clarification on unclear itemsin 68 of 668, across 56 files
- Verify feedback against codebase before implementationin 66 of 668, across 55 files
- Implement fixes one at a timein 64 of 668, across 53 files
- Test each fix individuallyin 62 of 668, across 51 files
- Restate technical requirements in own wordsin 57 of 668, across 46 files
- Reply to inline comments in the specific threadin 51 of 668, across 40 files
- Note minor issues for laterin 49 of 668, across 34 files
- Group findings by severityin 48 of 668, across 47 files
Said here and by no other author read
- Write comments in one line
- Use full explanation for security findings
- Use full explanation for architectural disagreements
- Use full explanation for onboarding contexts
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.