Link in bio for an agentic OS system

My notes

Summary

A short-form walkthrough of structuring a personal agentic OS around Claude Code, using Obsidian for memory and GWS-CLI for productivity. The key idea is organising AI skills like an org chart: top-level domains (sales, marketing, admin) each containing modular sub-skills and skill packs (higher-order skills that call other skills).

Key Insight

• The “org chart” mental model maps directly to how Claude Code skill systems work: domains as branches, individual skills as leaf nodes, skill packs as mid-level managers that orchestrate leaves.
• Memory (Obsidian) and productivity tools (GWS-CLI) are treated as infrastructure, not skills - they underpin everything else without being domain-specific.
• “Skill packs” = composite skills that call other skills; this is the correct abstraction for avoiding prompt duplication and making the system composable.
• The dashboard / command centre layer (slash-command interface) sits between the human and the agent graph - useful for discoverability when the skill count grows beyond ~10.
• Content and research are cited as the two highest-leverage skill categories for this creator’s workflow - suggests the pattern generalises to knowledge workers with similar bottlenecks.