This is the deprecated v1 version that runs a Stop hook at session end to extract reusable patterns from your Claude Code conversations and save them as learned skills. It checks if your session hit a minimum message count, identifies patterns like error resolutions or user corrections, and writes them to your learned skills directory. The v2 version is what you actually want now since it uses PreToolUse/PostToolUse hooks for 100% reliable observation instead of probabilistic skill firing, plus it breaks learning into atomic instincts with confidence scoring rather than full skills. This v1 is kept around for backward compatibility and archival reference, but route new continuous learning work to v2.
npx -y skills add affaan-m/ecc --skill continuous-learning --agent claude-codeInstalls into .claude/skills of the current project.
Select a file.
juliusbrussee/caveman
mattpocock/skills
obra/superpowers
forrestchang/andrej-karpathy-skills
vercel-labs/skills