Skip to main content
An App Definition is a document with a published schema and an offline checker. That’s a good deal for an AI agent. It can write something, check its own work, and be told exactly what it got wrong, with no database, no model API and no deploy anywhere in the loop. This section has a starting prompt for each of the common agents. The prompts differ because the tools differ, but the reasoning behind all of them is the same, so read this page first.

​
Your workspace already briefs the agent

That writes CLAUDE.md and AGENTS.md into the workspace root, same generated text in both. Two conventions, two filenames. Both get written because a workspace that taught two agents different rules would end up with source only one of them could maintain.
This is the most important thing on the page. The instructions are already there. Your prompt doesn’t need to re-explain Cordango, and it shouldn’t try.
What those files teach is the workflow: how to find things out. They deliberately don’t paste the vocabulary in. That’s what keeps an agent’s context from filling up with 212 KB of JSON Schema.

​
The rules they install

Every prompt in this section leans on these, so they’re worth reading once. Discover, don’t guess. cordango inspect for what exists, cordango vocabulary for what you’re allowed to write. Both answer from the installed build, so they can’t drift from what the CLI will accept. Never read the schema, never read the binary. If cordango vocabulary can’t answer a question, we’d rather hear about the gap than have an agent work around it. Check constantly. cordango check calls no model and touches no database. It’s free. Run it after every change. One aggregate per change. Diffs stay reviewable, and it’s what --scope enforces anyway. Stable keys. You can’t rename an entity or field key, because nothing rewrites the references to it. Add and remove deliberately instead. The agent doesn’t accept its own work. cordango apply writes to the working tree. A person reads git diff and stages it.

​
Every command speaks JSON

--json works on all of them. cordango help --json is how an agent discovers the command surface without a human pasting a README into its context.

​
The shape of a good prompt

Short, and about the job. The workspace supplies the rest.
Look at what’s missing. No explanation of what Cordango is, no YAML example, no schema excerpt, no list of field types. All of that is one cordango vocabulary field away, and an agent that looks it up gets the version matching the CLI it’s actually running.

​
Pick your agent

Claude Code

Reads CLAUDE.md automatically. The path of least resistance.

Codex

Reads AGENTS.md. Same workspace, same instructions.

Cursor

Needs a rules file pointing at the workspace instructions.

GLM

Strongest with explicit, sequenced instructions.