AI · Technology · 2026-09-09 06:15:00
Toolproof is an independent measurement project for AI agent tooling, aggregating nine indexes that read 755,255 agent-skill listings from source, of which 8.7% (65,390) fail to load. The indexes cover Claude Code skills, subagents, plugins and marketplace repositories, tracking whether tools are still maintained, daily AI coding tool changes, SaaS starter kits, AI app builders, shadcn registry components, agent config conventions, answer-engine citations and document-extraction APIs. As of August 14, 2026, the largest catalogues describe their methods similarly but none claims anything is executed; stars, installs and search demand measure discovery, not whether a tool still works. That gap is why the first figure is a failure rate, and why Toolproof publishes one shared, pre-stated method with conflicts named.
Source: toolproof.kynth.studio
Auto-generated summary for reference only. RSS