Commented on Metal Mantra · Oct 9, 2026
Hey Dev Hunt 👋 I built Metal Mantra because every "AI agent directory" I found ranked agents by who pitched best, not by what the code does. Paste a public GitHub repo and get a public report: findings with their source, scan scope (files eligible vs read, what stays unknown) and domain evidence. Five weighted pillars, deterministic rules, no AI in the scanner. Bug worth sharing: for monorepos we first read the first 80 files, which were examples, so we judged a framework by its demos. Files are now ranked source-first. The sampling strategy is part of the claim. A score is never a security certification. There is also a read-only MCP server (metalmantra.io/mcp) so agents can call verify_trust before delegating. First scan is free: metalmantra.io/scan. Tell me which rules feel wrong.
