Custom agent skills, promoted from incubation in niftymonkey/claude once they've passed a portability review.
These skills work with any agent supported by the skills CLI — Claude Code, Codex, Cursor, and ~50 more.
# Browse what's available before installing
npx skills@latest add niftymonkey/skills --list
# Install everything
npx skills@latest add niftymonkey/skills
# Install a single skill
npx skills@latest add niftymonkey/skills --skill <name>By default the CLI installs at project scope (./<agent>/skills/). Use -g for global (~/<agent>/skills/).
- continue — Capture full session state to a
continue.mdfile so a brand-new conversation can resume exactly where this one ended. Useful when the context window is filling up, before stopping mid-task, or any time you want a clean hand-off to a fresh session. (history) - pr-feedback — Address code review comments on a GitHub PR systematically. Fixes code, pushes changes, and responds inline to every comment with either the fix or specific reasoning for dismissing. Uses
ghCLI. (history)
Each skill starts in niftymonkey/claude at status incubating. When it's matured and proven useful, it goes through a /promote-skill review — automated checks for personal paths, credentials, and domain references, plus a collaborative pass on implicit assumptions a stranger wouldn't share. Only skills that pass land here.
The full design history for every skill is preserved in history/. It's deliberately kept outside the skills/ tree so npx skills add ships only runtime content — installs stay lean while the archival record stays public.
MIT — see LICENSE.
Some skills include content adapted from mattpocock/skills (also MIT). Adapted files carry a visible attribution line at the top of the file.