@salve-software/aiworkers

AIWorkers

Central package for AI configurations. Instead of each project having its own CLAUDE.md written from scratch, AIWorkers centralizes everything that is generic. Each project inherits with a single line and only writes what is specific to itself.

How it works

Any project creates a CLAUDE.md with one line:

@../../packages/AIWorkers/CLAUDE.md

Claude Code automatically injects all AIWorkers context — rules, commands, and agent personas. Below the import, the project writes only what is specific to it.

What lives here vs in the project

	AIWorkers	Project
Commit standards	✅
Reusable slash commands	✅
Agent personas	✅
Permissions (settings.json)	✅
Domain context		✅
Specific stack		✅
Business rules		✅
Project-specific commands		✅

Structure

AIWorkers/
├── src/
│   ├── commands/          # User-invokable slash commands → docs/commands.md
│   ├── skills/            # Context-triggered skills (auto-invoked by Claude)
│   ├── agents/            # Agent personas → docs/agents.md
│   └── rules/             # Composable rules (one responsibility per file)
├── docs/                  # Decisions, patterns, and reference docs
├── scripts/
│   └── setup.sh           # Links src/ into .claude/
├── CLAUDE.md
└── settings.json          # Source of truth for permissions

Setup

Run once per machine:

./scripts/setup.sh

The script symlinks commands, agents, and rules from src/ into .claude/, and keeps settings.json in sync. No restart needed after running.

Reference

• Slash commands
• Agent personas

Requirements

• Claude Code with an active Pro or Team plan
• gh CLI installed and authenticated (required for /pr and /feature)