Agent context
Agent context — get oriented
The shared context that keeps people and AI agents building Cynario the same way. What this section is, who it's for, and how to use it — a curated landing into the project's working brain, not a full mirror of it.
This is the deepest, most behind-the-scenes layer of the docs. The rest of the site explains what Cynario is and how to use it. This section explains how Cynario is built — the shared context that keeps every contributor, human or AI, pulling in the same direction.
The idea in one line: the same notes a person reads to get oriented are the notes an agent reads to stay consistent. One context, two kinds of reader.
Who this is for
Two audiences, the same material:
- People joining the project — contributors who want the map before they touch the code: where things live, how decisions get recorded, what the conventions are.
- AI coding agents — the assistants that build alongside us. They read
CLAUDE.md,AGENTS.md, and the skills under.claude/on every task. Those files are summarized here so a human can see, in plain prose, the same rules the agents are running on.
It is not for someone who just wants to build a scenario — that’s the Guides. This is one level under the Concepts and Architecture: past how the product thinks and into how the project works.
What’s in here
A curated landing — the entrances, not the whole building. Four pages:
| Page | What it gives you |
|---|---|
| Get oriented (this page) | Why this section exists and how to read it |
| Codebase map | The load-bearing modules and the seams between them — where to look first |
| How we work | The knowledge-base temperature gradient, and how we develop in the open without leaking infra |
| Agent config | What CLAUDE.md, AGENTS.md, and the .claude/skills/* are for — the consistency layer |
Each page is a map with links, not a copy. When it points at a real file in the repo, follow the link — the file is the source of truth; this is the orientation.
How to use it
- New to the project? Read this page, then the Codebase map, then How we work. Twenty minutes and you know where everything is and why it’s arranged that way.
- Picking up an agent’s work? Start with How we work — the knowledge-base gradient is where the reasoning trail lives, so you can continue the same direction instead of guessing it.
- Wondering why an agent did something a certain way? The answer is usually in Agent config: the rules it follows are written down, not improvised.
Why it’s published at all
Cynario is built in the open, so this working context is open too — but deliberately. Two axes keep it safe by construction:
- Machine and personal specifics are config, never content. Hosts, IP
addresses, and personal paths live in gitignored config; in prose we always
write placeholders like
<host>.<tailnet>.ts.net,<tailnet-ip>, and<repo-root>. An automated guard rejects any commit that smuggles a real one in. - Everything else is content that matures. Nothing here is secret; raw work just isn’t done. We publish what has cooled and settled and hold back what’s still raw.
Both axes are spelled out in How we work. The upshot: you’re reading the project’s actual working brain, with the machine-specific bits swapped for placeholders.