| name | rulesync-preferences |
| description | Use when adding, changing, or reorganizing AI guidance in this repository. Prefer Rulesync skills over commands, subagents, or hooks for reusable workflows, and always edit Rulesync sources before generating tool-specific AI config files. |
| targets | ["*"] |
Rulesync preferences
This repository treats Rulesync as the AI configuration system of record.
Preferences
- Prefer Rulesync skills for reusable AI workflows and agent guidance.
- Prefer Rulesync skills over commands or hooks for reusable workflows that should be loaded in-context.
- Prefer Rulesync subagents when the work is better handled by a delegated specialist with its own context window.
- Prefer creating or editing Rulesync source files over editing generated AI config files directly.
Workflow
- Decide whether the request is durable AI guidance rather than a one-off code change.
- If it is a reusable workflow, implement it as a skill under
.rulesync/skills/<skill-name>/SKILL.md.
- Edit
.rulesync/ sources only.
- Run
npx rulesync generate.
- Review generated outputs such as
AGENTS.md, .codex/*, .cursor/*, .claude/*, and .github/*, but do not patch them directly for durable changes.
Decision rule
- Choose a skill first when the guidance should help future agents converge on intent across repeated tasks and should be loaded directly in the active agent context.
- Choose a subagent when the work should be delegated to a specialist that can operate with its own prompt and context window.
- Avoid commands for durable reusable workflows in this repository.
- Use a rule only when the guidance must always apply ambiently across the repo rather than as a task-specific workflow.