skill-authoringlisted

# Skill Authoring Use this skill to create or maintain Agent Skills that load reliably and stay portable across Codex and Claude. ## Scope - Use for `SKILL.md` creation, review, and refactoring. - Use for package manifest updates when a skill should ship. - Use for discovery-trigger tuning when a skill is not invoked or triggers too broadly. - Do not use for product plugins, MCP servers, or runtime code unless the skill instructions need supporting files for them. ## Current Format Baseline - A skill is a directory with one `SKILL.md`. - `SKILL.md` starts with YAML frontmatter containing `name` and `description`. - The description is always-loaded metadata. Keep it concise, specific, and front-loaded with trigger words. - The Markdown body loads only after the agent selects the skill. - Put long examples, API notes, templates, and scripts in supporting files such as `references/`, `assets/`, or `scripts/`, and point to them from the body. - Codex uses the open Agent Skills format and discovers repository/user skills from `.agents/skills` and `$HOME/.agents/skills`. This package still has a legacy campaign runtime config path under `.codex/skills/project.toml`; do not rewrite that runtime path casually. ## Workflow 1. Decide whether a new skill is warranted: - repeated workflow, checklist, or project convention - long prompt that should be reusable - specialized procedure that benefits from on-demand loading 2. Pick one narrow name: - kebab