agentsclimarketplace
Sign in

Ultrareview

Skill NafisRayan/100x-Agent-Toolkit/skills/ultrareview

Multi-agent parallel code review system inspired by Claude Code's /ultrareview. Spawn 5-20 specialized agents (security, logic, performance, edge cases, architecture) to review code simultaneously, cross-validate findings, and produce a consolidated bug report. Use when asked to do deep code review, ultrareview, multi-agent review, comprehensive bug hunting, security audit, or thorough pre-merge code inspection.From its SKILL.md

Install
npx -y skills add NafisRayan/100x-Agent-Toolkit --skill ultrareview

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

  • no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
  • 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

4.2 KB, 898 tokens by cl100k_base, as published. Nobody here has run it

Ultrareview

Multi-agent parallel code review: spawn a fleet of specialized review agents, cross-validate findings, and output a consolidated report with severity levels.

Quick Start

  1. Determine review scope (files, directories, PR diff, or full project)
  2. Spawn 5 parallel review agents (scale to 20 for large codebases):
    • Logic Verifier — algorithmic correctness, off-by-one, unreachable code
    • Security Sentinel — OWASP Top 10, injection, auth bypass, secrets
    • Performance Oracle — N+1 queries, blocking I/O, complexity
    • Boundary Inspector — null handling, empty collections, overflow, timezones
    • Architecture Reviewer — SOLID violations, coupling, API mismatches
  3. Cross-validate: each agent challenges other agents' findings; discard unconfirmed issues
  4. Aggregate, deduplicate, rank by severity, output report

Review Process

Phase 1: Context Collection

  • Read project's CLAUDE.md and REVIEW.md for custom rules
  • Identify scope: specific files, git diff, or full codebase
  • Map entry points, dependencies, and data flow paths

Phase 2: Parallel Agent Dispatch

Spawn agents as subagents or Agent Team members. Each agent receives:

  • Full code context for the review scope
  • Specialization prompt defining its review angle
  • Project-specific rules from CLAUDE.md / REVIEW.md

Minimum 5 agents for meaningful cross-validation. Scale up for large/critical codebases.

Phase 3: Cross-Validation

After initial findings:

  1. Share findings across agents
  2. Each agent attempts to disprove others' findings (check if issue is handled elsewhere, verify fix won't regress)
  3. Retain only findings confirmed by evidence
  4. Assign confidence score based on cross-agent agreement

Phase 4: Report Generation

Severity classification:

  • 🔴 Critical: Must fix. Confirmed bugs, security vulnerabilities, data corruption
  • 🟡 Warning: Should fix. Potential issues, performance concerns, code smells
  • 🟣 Pre-existing: Not introduced by current changes. Historical tech debt

Each finding includes:

  • File + line range
  • Issue description
  • Root cause analysis
  • Suggested fix with code
  • Reasoning trace (collapsible)

Output Format

# Ultrareview Report
**Scope**: [files/directories reviewed]
**Agents**: [count] | **Duration**: [time] | **Findings**: [count]

## 🔴 Critical (N)
### [C1] [Short title]
- **Location**: `path/to/file.ts:42-58`
- **Issue**: [Description]
- **Root cause**: [Why this is a bug]
- **Fix**: [Suggested code change]

## 🟡 Warning (N)
### [W1] ...

## 🟣 Pre-existing (N)
### [P1] ...

## ✅ Areas Reviewed — No Issues
- [List of clean areas]

Excluding From Review

Respect REVIEW.md ignore patterns. Common exclusions:

  • Generated code, vendored dependencies, test fixtures
  • Style/formatting issues (never flag these)

Implementation Options

Choose based on your environment:

Architecture Deep Dive

For execution flow diagrams, agent specialization details, cross-validation mechanics, cost benchmarks, and product comparison table → see references/architecture.md.

What ships with it: 3 files

21.5 KB alongside SKILL.md

references/

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.