Agent Skills

A small collection of reusable agent skills.

Quickstart

Install all skills:

bunx skills add zanellig/skills

Install selected skills:

bunx skills add zanellig/skills --skill bump --skill commit

Install from a local checkout:

bunx skills add ~/projects/skills --skill bump --skill commit --full-depth

Explicit-only skills

A skill that should never fire on its own — one that spends usage, writes to version control, or schedules recurring work — has to say so twice, because each harness reads its own key:

# SKILL.md frontmatter — Claude Code
disable-model-invocation: true

# agents/openai.yaml — Codex
policy:
  allow_implicit_invocation: false

Either one alone leaves the skill self-selectable in the other harness. Both keep it out of the model-visible catalog while /skill-name and $skill-name still work.

To check what Codex actually sees, without spending a request:

codex debug prompt-input | grep -A200 skills_instructions

Reference

See skills/README.md for the full list, generated from each skill's frontmatter:

bun scripts/gen-reference.mjs

The same command updates generated skill-reference sections in existing category and subcategory README.md files. It preserves content outside the generated markers and does not create new category READMEs.

You should not need to run it by hand. Enable the pre-commit hook once per clone and it regenerates and stages the reference on every commit:

git config core.hooksPath .githooks

CI runs bun scripts/gen-reference.mjs --check and fails if the committed reference is stale, so this stays a backstop rather than the thing you rely on.