One Shared Skills Folder Made Every Agent Portable
Instead of letting every agent create its own skills directory, one shared folder and a symlink setup made skills portable across apps.
The Problem
Each agent was creating its own skills directory, which meant useful capabilities were getting fragmented and trapped inside different applications.
What Was Built
A single shared AI skills folder was created, then each agent-specific skills directory was symlinked to it. From each app's perspective, it still looked like it had its own local skills folder, but in practice everything pointed to the same shared source of truth.
Where AI Sits in the Workflow
AI helped identify the portability problem and design a cleaner shared structure for how skills should be stored and reused. A person decides which skills are worth sharing broadly and updates them when the underlying process changes.
Tools Used
The Result
Skills no longer have to be recreated or manually copied across applications, which makes the whole AI stack easier to maintain.
Key Insight
A lot of AI workflow friction is not model capability. It is file-system fragmentation.
Want this built for your business?
Want a workflow like this in your business? Talk to Dovid.
Start the conversationMore Examples
An Agentic Doc Repo Makes Documentation Retrievable
Instead of living like normal docs in Notion or Confluence, the repo is designed so agents can actually retrieve what they need.
A Vectorized Memory Layer Gives Agents Shared Context
Important work context can be written once, stored in memory, and read back by agents through an MCP server later.
A Daily Agent Turns Conversations Into Memory
Every day, a scheduled agent reviews Codex conversations, identifies what work is happening, and turns that into usable memory automatically.