Pr reviewer
Always-on multi-agent AI workstation for Claude Code. 142 specialist agents (ML, security, K8s, backend, frontend) · 366 curated skills · 200+ prompt templates · auto prompt-matcher · auto skill-matcher · MCP support · token router cuts Anthropic bill 6x via Kimi/Haiku/Ollama. LiteLLM · BYOK · MIT + Apache. One-line install.From the repository description
npx -y skills add timdevai/proteus --skill pr-reviewerAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
2.9 KB, 762 tokens by cl100k_base, as published. Nobody here has run it
pr-reviewer
Pre-reviews your PR before you push. Diffs against main, flags risks, checks test coverage delta, suggests reviewers.
When to activate
Trigger phrases: review my PR, pre-pr, before I push, am I ready to push, /pr-review.
Auto-activate when user runs git push on a feature branch (PreToolUse hook, optional install).
Process
- Determine the base branch (
main,master, or whatever GitHub default is) - Compute the diff:
git diff <base>...HEAD - Run checks below
- Generate a review report with risk score
- Suggest fix-before-push items vs nice-to-have
Checks
Risk flags (each = -10 to safety score)
- Force push detected
.env,*.key,*.pemin diff- Hardcoded secrets pattern (
sk-,Bearer,AKIA,ghp_) -
500 line single-commit diff (suggest split)
- Migration file added but no rollback script
- New endpoint with no auth check
- Direct DB call in route handler (bypasses repository pattern, if used)
- New external dependency added (which one, why, alternatives)
Yellow flags (each = -3)
- No new tests added but new production code present
- TODO/FIXME added without ticket reference
- Console.log / print() / debug statements in diff
- Commented-out code blocks
- File renamed without rg-checking for references
Green flags (each = +3)
- Tests added that cover new code paths
- ADR linked in PR body
- Conventional commit messages used
- Diff < 200 lines
- README/docs updated alongside code
Output
PR Review: feat/rate-limit-middleware → main
=============================================
Diff stats: 6 files changed, 142 +, 18 -
Safety score: 78/100
✗ Risk (blocks push):
none
⚠ Yellow flags:
- 1 TODO added without ticket: middleware/rate_limit.py:45
- console.log left in: middleware/rate_limit.py:67
✓ Green flags:
- Tests added (test_rate_limit.py, 4 cases)
- Diff is 142 lines — focused
- Conventional commits used
Suggested reviewers: (based on git blame)
- [email protected] (touched middleware/ 12 times this quarter)
- [email protected] (owns auth module)
Suggested PR body:
## What
Adds IP-based rate limiting to /login.
## Why
Prevents brute-force after 5 failed attempts in 60s.
## Test
pytest tests/test_rate_limit.py
## Risk
Low — fail-open if redis unreachable.
Fix before push:
1. Remove console.log on line 67
2. Either delete TODO or link to a ticket
Ready to push? (y / fix-then-push)
Inputs
- Current branch + base branch
- Diff
- git blame for "suggested reviewers" feature
Outputs
- Review report
- Suggested PR body
- Fix list
- Reviewer suggestions
Slash command
/pr-review — run on current branch.
/pr-review push — run review, then push if score ≥75 with no red flags.
Why this skill
Self-review catches 60% of "junior" mistakes before another human spends time on it. Pre-push is the cheapest place to fix anything.
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.