SkillWorkshop

Six tools distilled from coding agent engineering: memory, compression, verification, routing, and orchestration.

Three years of Andrej Karpathy’s public work distilled into 14 installable Agent Skills, covering research loops, idea files, LLM debates, reflection, and coding principles.

Claude Code handles planning and review; Codex handles implementation and completion. A Session Receipt records the work.

A macOS menu bar sound beacon: hear when an agent starts, finishes, needs permission, or gets interrupted.

Analyze a creator’s successful content without an API and distill their approach into an installable content coach.

Turn a Feishu document or local Markdown into an X Article draft, with media restored to its original positions.

An action-oriented WeRead reading coach: choose a section to read today based on the problem you are trying to solve.

A daily watch on ClawHub’s Top 100, tracking which Skills are gaining momentum.

A collection of early LLM examples and guides, documenting where it all began.

Yugong: give an agent a loop that keeps making progress, without manually prompting every round.

Cailun: turn useful ideas from a conversation into a shareable page instead of leaving them buried in chat history.

Afu: plan actions with LLM cards, combine related tasks, sync calendars, and see what to do today.

A collection of real AI workflows for creators and a central index of related skills.

A ready-to-use collection of Irasutoya-style illustrations for articles.