Build a context bundle before asking AI to code

I'm about to ask an AI to write code for the task below. Help me build the context bundle that grounds the request, so the AI doesn't have to invent APIs it doesn't know.

The task:
{{describe what you want the AI to build, 1-3 sentences}}

What surfaces this code touches:
- Database tables / schema: {{which tables, or "none"}}
- External APIs / libraries: {{Stripe, OpenAI, internal libs, etc.}}
- Internal modules it'll call: {{file paths or function names}}
- Types / interfaces it'll need to match: {{TypeScript types, Zod schemas, protobuf, etc.}}

For each surface, tell me:
1. **What to paste** — the exact source-of-truth I should include with the task
2. **What to leave out** — irrelevant context that'll just dilute the prompt
3. **What to call out explicitly** — invariants, naming conventions, gotchas the AI won't infer from code alone

Then assemble it into a single context block I can paste *before* the actual task description. Format:

```
## Context

### Schema (relevant fields only)
[...]

### Library APIs
[...]

### Internal functions
[...]

### Conventions
- [invariant]
- [invariant]
```

If a surface needs context I haven't given you, ask before assembling.