skill-authoringlisted

# Skill authoring conventions Rules for authoring or editing skills in `~/.dotfiles/claude/skills/`. Apply to every skill, present and future. ## Autodetection over named cross-references Skills load based on their own descriptions matching the project context, not because one skill instructs Claude to load another. A skill's description must: - Trigger on concrete project signals (file extensions, sentinel files like `manage.py` or `next.config.js`, dependency markers in `package.json` or `pyproject.toml`, distinctive syntax). - Trigger on keyword aliases (the user mentioning the technology, even informally). - NOT instruct Claude to "load skill X alongside this one" or "see also skill Y". - NOT enumerate companion skills in the body as "load these together". Skills load in combination because each skill's independent triggers all match the same project. A TypeScript Next.js project loads `typescript-patterns`, `react-patterns`, and `next-app-router-patterns` because each description independently matches the project's signals, not because one skill names the others. Documentation references to other skills as concepts ("framework-specific patterns live in their respective skills") are acceptable. Load instructions that name other skills are not. ## Description shape > [What the skill is, one phrase]. Use whenever [project signals], OR the user asks about [keyword aliases], even if [the technology] is not mentioned by name. Both project signals and keyword aliases a