Governances Overview
Governances are version-pinned, auditable agent-tool contracts shipped with the cyberplace npm package. They are frozen to the installed version and loaded on demand via the CLI.
Disciplines (e.g. commit discipline) are a separate layer — session-scoped rules injected by hooks. See Commit Discipline.
Loading governances
Section titled “Loading governances”Do not link to governance files directly from SKILL.md. Load them through the CLI so the content is always version-matched:
# List available governancesnpx cyberplace@<version> governance list
# Show a governance (agents read stdout)npx cyberplace@<version> governance show skill-design
# Agent-optimized output (lower token cost)npx cyberplace@<version> governance show skill-design --format agentAlways pin an exact version: npx cyberplace@$(npm view cyberplace version) …
Available governances
Section titled “Available governances”| Name | Purpose |
|---|---|
| skill-design | Rules for authoring SKILL.md files |
| skill-repo-structure | Rules for organizing a skill library repository |
| agent-tool-output | Output rules for scripts, hooks, and CLIs that agents invoke |
| cli-resolution | Strategy for invoking a Node CLI that may be global, repo-local, or absent |
| universal-plugin | Format spec for plugins that work across Claude Code, Cursor, and Codex |
Authoring rules
Section titled “Authoring rules”Governances are agent-first — they load directly into agent context:
- Dense and concise — imperative must/should/do-not rules; no tutorials
- Self-contained — no links to other repository files; agent completes the workflow from stdout alone
- References at end — only
governance showcommands and HTTPS URLs in## References - No rationale sections — no
## Why,## Rationale, or## Background; ADRs record why, governances record what