The Context Collection is where the CAS open standard becomes a ready-to-use product — expert-built shells you install into any AI model and put to work in your business today.
Brand voice, content strategy, and prompt cards for on-brand marketing content.
78 cards, every spread, any question. Gives any AI model the knowledge to read tarot with depth and accuracy.
Structured context packs, reusable AI workflows, agent-ready knowledge systems, and interactive digital experiences — designed for use with modern language models.
Organized collections of prompts, instructions, frameworks, rules, reference material, and workflows that help ChatGPT, Claude, Gemini, local models, and future agents produce more consistent and useful results.
Reusable context products for a world where intelligence is interactive, portable, and personal.
Purchase and download. Create a private GitHub repository. Run the setup script. Two minutes.
Add your brand voice, your offers, your existing copy. The shell guides you through what it needs. Start with what you have.
Paste the setup prompt into Claude, ChatGPT, Gemini, or any model. It reads your context and gets to work. No rebriefing. No starting over.
Every shell here is built on the Context Architecture System™ — a free, open standard for portable AI context. These ten principles govern what CAS commits to and what it refuses to compromise on.
The context layer belongs to whoever's work it describes. Not to a vendor, not to a model provider, not to a consultant. Ownership is the first commitment; every other principle follows from it.
Plain Markdown, plain folders, plain text. No proprietary format. No database. No app required to read it. Portability is the test — a context that cannot be moved between machines, services, or decades is not a context worth keeping.
Git provides history, attribution, and replication without requiring a service. Where Git is not an option, the context is still valid. Where Git is in use, every meaningful change should produce a commit.
Models will be replaced. Vendors will change pricing or shutter. Consultants will move on. The context must remain readable and usable through all of these transitions.
The default permission for any AI model or agent is read-only. Changes must be proposed, reviewed, and applied by the owner — not made silently. Destructive operations require explicit human instruction.
A change worth making is a change worth recording. Skipping commits for trivial edits is fine; skipping them for substantive ones surrenders accountability to memory. The git log is the durable record.
API keys, passwords, tokens, and credentials belong in a password manager — not here. A context that contains secrets becomes a liability the moment it is shared, backed up, or pushed to a host.
Every account referenced in the context is registered in the owner's name. No account is held by a consultant or third party on the owner's behalf. When the relationship ends, nothing of value leaves with them.
Any tool, host, service, or runtime can be swapped for an equivalent without rewriting the context. If switching vendors would require structural changes to the context, the infrastructure is too tightly coupled.
The point is not the files themselves. The point is that work continues — through model changes, vendor changes, team changes, and time. A context system that survives ten years is the goal.
Every shell in The Context Collection is built on the Context Architecture System™ — a free, open standard for portable AI context. Your files are yours. They work with any model, any tool, any time.
Learn about CAS →