Skip to content
github-actions[bot] edited this page Jul 1, 2026 · 5 revisions

AI Layer

How agentic-workstation provisions AI skills, agents, prompts, and templates.


Overview

AI resources are deployed to ~/.local/share/agentic-workstation/ during chezmoi apply:

Directory Contents
prompts/ Reusable internal prompts
skills/ Bundled skills (managed by chezmoi) + skill-catalog.yaml
skills-external/ External skills installed by dots-skills
templates/ Reusable text templates
mcp/ MCP provider examples and wrappers
skills-registry.yaml Runtime registry for skill management

Skills

Skills teach AI tools how to perform specific workflows. Each skill has:

  • SKILL.md — instructions read by AI tools
  • skill.json — manifest with compatibility matrix

Skills are symlinked to each AI tool's config directory by dots-skills sync:

Tool Skills directory
Claude Code ~/.claude/skills/
OpenCode ~/.config/opencode/skills/
Cursor ~/.cursor/skills/
Copilot CLI ~/.copilot/skills/

→ Full details: Skills System


Agents

15 specialized subagents are installed to tool-specific directories:

  • ~/.claude/agents/ (Claude Code)
  • ~/.config/opencode/agents/ (OpenCode)
  • Cursor and Windsurf rules directories

Use @agent-name to invoke in Claude Code or OpenCode.


The Ralph Loop

The workstation implements a Ralph Loop pattern:

Ralph concept What the workstation provides
Backing specifications AGENTS.md templates deployed to each repo/session
Context engineering Skills that prime each loop with domain context
Persistent memory ai-workspace/knowledge/ knowledge base
Fix the loop dots-harness-knowledge-sync skill
Monolithic orchestrator dots-workstation-assistant as single entry point

Canonical doc: docs/AI_LAYER.md

Clone this wiki locally