author-skillslisted

# Advanced Skill Authoring Self-contained — nord's single home for skill authoring (foundation + advanced patterns). ## 0. Foundation (adopted from writing-skills) The basics this skill builds on (grafted so it stands alone): - **A skill = a focused capability**: `SKILL.md` with YAML frontmatter + a concise body. Selection is decided by `name` and `description` alone, optionally extended by `when_to_use`. **`triggers:` is not a Claude Code field** — the harness ignores it, and several older skills in this tree carry one that has never done anything. Description = when it fires (symptoms), NOT a workflow summary (else the body is skipped). - **Name it after the question the user arrives with**, not the technique inside — `plan`, not `roundtable`. The wider convention is gerund-first (`writing-plans`); nord-core's own catalogue uses the bare verb, for the reasons in vault `decisions/skills-nach-fragen-statt-nach-mechanismen.md`. - **Concise output (CSO).** Body is reference the model reads under load — terse, scannable, no filler. Tables/lists over prose. Every line earns its tokens. - **TDD for skills (the Iron Law).** Before trusting a skill, pressure-test it: write 3+ representative prompts, run in a fresh session, confirm the right behavior fires; adversarially test that adjacent prompts don't mis-trigger it. A skill with no eval evidence is a hypothesis, not a validated tool. - **Plugin distribution**: a plugin's skills live in `<plugin>/skills/<name>/S