Rico's Tinfoil Hat

A public, research-backed ranking of the highest-leverage Claude Code skills and skill systems, with links to the GitHub repositories or official docs where developers can install or learn them.

A Reddit user accidentally named the most useful Fable 5 workflow: use the strongest model while you still have it to write skills, prompts, memory rules, and harness instructions that make cheaper models behave more like it later. It is not magic model transfer. It is operational capture.

Claude Sonnet 5 is not the wildest model Anthropic has shipped this month. That is exactly why it matters. After Fable 5 and Mythos 5 were pulled under an export-control directive, Sonnet 5 looks like the practical agent model developers can actually use: cheaper than Opus, much stronger than Sonnet 4.6, and built for long-running tool use.

Yesterday's Code with Claude keynote was not a Claude 4 launch. Anthropic's real May 6, 2026 announcements were about Managed Agents getting dreaming, outcomes, and multiagent orchestration, plus higher Claude Code limits, finance-focused agent templates, and a stronger push toward long-running autonomous workflows.

OpenClaw went from 60K GitHub stars to selling out Mac Minis nationwide, then its creator joined OpenAI. Perplexity launched a $200/month multi-model agent platform. Anthropic shipped phone-to-PC coding with Remote Control. The fad pointed at the future: AI that operates, not assists. Built on RAG, markdown, and skills. This is where it's all heading.