Show HN: AI vs. AI – code and reviews only count if they survive an attack

Category: devtools

Tags: llm-evaluation, adversarial-testing, code-generation

Score: 6.5/10 (Innovation: 6, Technical: 7, Documentation: 8, Utility: 5)

Pantheon-Skills is a multi-agent harness for Claude Code that runs coding tasks through plan, parallel implementation, adversarial verification, and judge stages to catch bugs that a single pass would miss. It's interesting because it packages well-known techniques into a reusable command, with a cross-model variant that uses GPT-5.5 to break single-model blind spots, and a separate skill for gap analysis on existing projects.

Target audience: backend devs, devops

Repository: https://github.com/lolu1032/pantheon-skills · JavaScript · MIT

View on Hacker News