This repository is the single source of truth for the AI-agent assets shared across the Spine SDK organisation: skills, scripts, and guidelines used by Claude Code, Codex, Junie, and other compatible agents.
Consumer repositories don't copy these files. They mount this repo as a
floating Git submodule at .agents/shared (tracking master) and expose it
through symlinks — .agents/skills, .agents/scripts, .agents/guidelines, and
.claude/commands, .claude/agents, .claude/skills. The
config repository wires this up
through its pull / adopt-shared-agents scripts, which float every checkout to
the latest master. The upshot: a change here reaches every repo on its next
pull, with no file churn in any consumer's pull requests.
skills/ # one directory per skill: <name>/SKILL.md (+ agents/openai.yaml, references/, scripts/, assets/)
scripts/ # shared helper scripts invoked by skills and agent hooks
guidelines/ # shared prose guidance (coding, testing, safety, docs, …) + _TOC.md
claude/ # Claude Code wrappers: commands/ (slash commands) and agents/ (subagents)
docs/ # documentation about THIS repo (project.md and the authoring guide)
The same content is also exposed under .agents/ via in-repo symlinks, so a
skill's own repo-rooted references (e.g. .agents/guidelines/coding.md)
resolve while you work here — the repo dogfoods its own skills.
The main contribution here is creating or editing skills. Start with the
skill-authoring guide and use the
author-skill skill to scaffold a new one. The
general contribution process is in CONTRIBUTING.md.
Because consumers float to master, treat it as production: land changes through
reviewed pull requests.