The problem with generic templates in managing agentic workspaces
Imagine you are the technical lead of a team that develops multiple products, such as Odoo modules, an authentication tool and a branding system. To save time, your team duplicates a workspace template containing generic folders, skills and rules. At first it works, but soon each product needs different structures: Odoo needs a specific module architecture, branding requires brand assets, and authentication has its own workflows. Keeping a single template up to date for all of them becomes chaos. Nobody updates the master template, each member ends up creating their own versions, and teamwork suffers.
At Ganemo we faced exactly this problem. We had teams working on Odoo, on our authentication product and on thesis projects, all using the same generic template. In the end, each one customized their workspace locally, losing synergy and duplicating effort. That is when we decided to implement AWAC, Agent Workspace as Code, which allowed us to define per-product configurations, centralized and updated from a single source of truth.
What AWAC is (Agent Workspace as Code)
AWAC is a convention that treats agentic workspaces as code. Instead of copying static templates, you define your workspace configuration (skills, rules, workflows, folder structure) in a central repository, and an AI agent applies it automatically when creating a new project. This ensures each workspace gets exactly the assets it needs, no more and no less, optimizing tokens and speed.
For example, at Ganemo we have an AWAC repository with templates for each product: blenout (authentication), odoo, branding, thesis, etc. When someone needs a workspace for a new project, they simply tell the agent something like: "Set up this workspace using AWAC for the blenout product". The agent queries the repository, downloads the specific skills and rules, and in less than a minute you have a workspace ready to work in, with the folder structure, the agent skills and the appropriate workflows.
How to implement AWAC in your team
To implement AWAC, you need a Git repository (public or private) with the following structure: one folder per product or team, each containing a configuration file (for example, workspace.yml) that defines the skills, rules and folder structure. Then you use a CLI (such as the one we offer on GitHub under the MIT license) so that AI agents (such as Cloud, Cline or Antigravity) apply the configuration when creating a workspace.
Suppose your company works with Odoo and also with marketing projects. In the AWAC repository you would create two folders: odoo and marketing. In odoo/workspace.yml you would specify skills such as "odoo_module_creator" and rules for the module structure. In marketing/workspace.yml you would put content generation skills and branding rules. When a developer wants to start a new Odoo module, they only need to run the CLI pointing to the "odoo" product and the agent will set everything up automatically. If another team member improves a skill, it is updated in the repository and all future workspaces inherit it.
At Ganemo, we implemented AWAC first in our Odoo team and then extended it to others. Today, any Ganemo member, whether developing Blenout or working on a thesis, can create a custom workspace in seconds without relying on outdated templates.
Key benefits of AWAC
- Per-product customization: each workspace gets only the assets it needs, optimizing resources.
- Centralized updates: improvements to skills and rules are automatically reflected in all future workspaces.
- Team collaboration: everyone uses the same source of truth, avoiding diverging local versions.
- Speed: complete workspaces created in less than a minute.
If your team faces the same challenge of keeping templates up to date and you want to improve productivity, AWAC is the solution. We at Ganemo have validated it internally and with clients, and that is why we released it as open source. If you need help implementing it, we can advise you.