swarm

Orchestrate 25 read-only agents to audit gnubok code quality, security, compliance, and UX.

209|46|Updated Feb 13, 2026
One-click install
npx skills add https://github.com/erp-mafia/gnubok --skill swarm-erp-mafia
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: swarm
Source: https://github.com/erp-mafia/gnubok/tree/main/.claude/skills/swarm
Command: npx skills add https://github.com/erp-mafia/gnubok --skill swarm-erp-mafia

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

The swarm skill orchestrates multiple read-only audit agents to produce a comprehensive, deduplicated set of findings for a gnubok codebase, enabling rapid governance and ticketing planning.

Core Features & Use Cases

  • Parallel audit agents: Launches 25 domain and cross-cutting agents to review Swedish VAT, security, UI/UX, and more.
  • Unified findings: Produces a flat numbered findings list with agent-wise grouping and severity sorting.
  • Issue deduplication & ticketing: Compares findings against open GitHub issues and creates approved tickets.
  • Usage patterns: Execute all agents, subsets by domain or model, or specific operator groups.

Quick Start

Initiate a full swarm run with /swarm to audit the entire codebase and generate a consolidated findings report.

Frequently Asked Questions about swarm

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a parallel code audit across an entire repository?

This approach launches 25 domain and cross-cutting agents, including Opus and Sonnet models with varied scopes, to perform a parallel evaluation across your entire repository, aggregating the results into a single consolidated dataset for ticketing decisions.

How do I deduplicate code audit findings against open GitHub issues?

Deduplicate audit findings by comparing the generated flat numbered findings list against your existing open GitHub issues, ensuring only new, approved tickets are created for unresolved problems.

What is a multi-agent audit workflow for Swedish accounting software?

A multi-agent audit workflow for Swedish accounting software coordinates specialized read-only agents to review domain-specific rules like Swedish VAT, alongside cross-cutting security and UI/UX evaluations, producing a deduplicated findings list for governance.

Can I audit specific domains instead of running a full repository evaluation?

Yes, you can audit specific domains by executing subsets of agents filtered by domain or model, or by targeting specific operator groups, rather than initiating a full swarm run across the entire codebase.

Do I need to worry about agents modifying my code during an automated audit?

No, the automated audit enforces strict non-editing rules by operating as a read-only evaluation, ensuring your codebase remains untouched while agents inspect code quality, security, compliance, and UX.