Minimal configuration so AI coding assistants can operate on this repo without rediscovering the build/release flow each session.
/build— build the CLI, the SDK bridge, or the release installer. Playbook wraps build.md./release— cut and push a SemVer tag, then watch the release workflow. Playbook wraps release.md.
Contributing rules (commits, branches, PR format) live in ../CONTRIBUTING.md.
| Path | Purpose |
|---|---|
../CLAUDE.md |
Project identity + hard constraints. Always in context — keep it short. |
../.claude/settings.json |
Read-only command allowlist (git status, gh run list, …). Reduces permission prompts without granting write access. |
../.claude/commands/*.md |
Slash commands. Loaded on demand when the user types /<name>. |
../.claude/skills/<name>/SKILL.md |
Skills. Auto-load when the description: matches the user's request. |
- Command — drop
.claude/commands/<name>.mdwith adescription:frontmatter line. Invoke with/<name>. - Skill — create
.claude/skills/<name>/SKILL.md. Auto-loads based on the trigger description (e.g. "when editing files that importfoo"). Prefer a command unless the knowledge should apply without being explicitly invoked. - Subagent — create
.claude/agents/<name>.md. Only worthwhile when a task recurs often enough to justify an isolated sub-context.
Rule of thumb: slash invocation → command; auto-apply while editing certain code → skill; own context window → subagent.
- Hooks — no confirmed "every time X → do Y" workflow yet.
- Subagents — no task recurs often enough yet to justify an isolated sub-context.
Not configured.