Knowledge card L2 · Context engineering informational

Put Reference Files in the Project, Not Every Prompt

What it is

A setup habit: upload the documents a body of work depends on — specs, guidelines, data — into the Project itself, so every chat inside it can draw on them without you re-pasting.

upload once, reuse everywhereEvery chat inherits both3Project instructions2Reference docs uploaded once1
Files in a Project are shared context for every chat inside it.

Why it works

Re-attaching the same reference material to prompt after prompt is wasteful and error-prone. Project files become shared context across every chat in that Project, so the model can reference your spec or brand guide in any thread without you reminding it — the context lives with the work instead of the message.

When to use it

Any multi-chat effort with stable reference material: a client with brand guidelines, a codebase with a spec, a research project with core papers.

When not to use it

One-off tasks with no ongoing context, or files relevant to just a single question — attach those to the prompt, not the Project.

Prompt

[In a Project] I've uploaded <the reference docs> to this Project. In every chat here, treat these as authoritative context — follow the spec/guidelines in them without me re-pasting each time.

Example

Uploading a style guide and product spec into a Project means every new chat drafts on-brand, on-spec copy without you re-explaining the rules.

Common mistakes

  • Re-pasting the same reference doc into every prompt.
  • Putting files in the Project but never telling the model to treat them as authoritative.
  • Dumping unrelated files into one Project so the shared context gets noisy.

Related