Caleb's personal agent skills library — reusable instruction sets for Claude Code and other agents, spanning engineering, church work, productivity, and whatever else earns a skill. Anything scoped to a specific project lives in that project's repo instead.
Uses the skills CLI, which installs SKILL.md files into whichever agents you target.
# Install from this repo (interactive — pick skills and agents)
npx skills add calebwaldner/skills
# See what's in here without installing
npx skills add calebwaldner/skills --list
# Install one skill, globally, to Claude Code
npx skills add calebwaldner/skills --skill design-handoff -g -a claude-code
# Install everything, everywhere, no prompts
npx skills add calebwaldner/skills --allScope is per-project by default (.claude/skills/); -g installs to ~/.claude/skills/ so the skill is available across all projects. Symlinking is the default and recommended install method — it keeps a single canonical copy, so npx skills update picks up changes here.
npx skills update # update installed skills to latest
npx skills list # show what's installed
npx skills remove design-handoff| Skill | Category | What it does |
|---|---|---|
design-handoff |
engineering | Compacts a grilled feature thread into a self-contained Claude Design handoff doc with a two-round prototype ask. |
delegate |
engineering | Delegate coding work to subagents. |
delegate-to-agents |
engineering | Staffs delegated work with specialists from The Agency catalog, installing missing agents after a yes, then hands execution to delegate. |
work-local |
engineering | Work on coding tasks locally without committing changes, allowing for review and manual commit. |
skills/<category>/<skill-name>/SKILL.md
One directory per skill. The category directory is organizational only — the CLI discovers skills at any depth, so grouping is free.
npx skills init skills/<category>/<skill-name>Then fill in the SKILL.md. See CLAUDE.md for the conventions this library follows — frontmatter, the Steps/Rules/Template shape, and why the rules are written to explain themselves.
Ideas that aren't skills yet live in IDEAS.md — a someday/maybe list with notes and open questions per idea.