> ## Documentation Index
> Fetch the complete documentation index at: https://docs.superscale.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# How context is used

> Understand what the agent uses now, what stays in the current chat, and what becomes durable context.

Uploading something, attaching something, and teaching Superscale something for the future are different actions. This distinction matters because it decides what the agent uses for one task versus what it remembers across future work.

The simple rule: **if a fact, rule, image, or learning should affect future generations, save it as scoped context.** If it only matters for the next output, keep it in the task prompt or attach it to the current chat.

## The context states

| State                       | What it means                                                                                       | Use it for                                                                                              |
| --------------------------- | --------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- |
| **Chat-only**               | The agent can see it in this conversation, but it is not automatically a future rule.               | A temporary campaign brief, one-off promotion, quick screenshot, or experimental instruction.           |
| **Used in this generation** | The agent actively uses it for the current output because it was selected, attached, or referenced. | The product, winning ad, reference image, script, or asset that should shape this creative.             |
| **Saved to product**        | The agent can reuse it whenever that product, offer, app, collection, or landing page is selected.  | Exact packaging, app UI, product claims, offer details, SKU rules, or funnel context.                   |
| **Saved to brand**          | The agent can reuse it across every product under that brand.                                       | Voice, logo rules, colors, approved claims, forbidden claims, audiences, competitors, and visual style. |
| **Saved to workspace**      | The agent can reuse it across every brand in the workspace.                                         | Agency QA rules, naming conventions, approval process, legal review rules, and team standards.          |
| **Saved learning**          | A result from research, reporting, or performance gets turned into reusable context.                | "This hook works for new parents", "avoid studio-style visuals", or "this format fatigues quickly."     |
| **Stored asset**            | The file exists in your workspace, but may not be used unless selected, scoped, or referenced.      | Old exports, source footage, backup logos, unused screenshots, or general asset folders.                |

<Note>
  Stored is not the same as active. If an asset or rule must guide a generation, select it, attach it, reference it in the prompt, or save it to the right context level.
</Note>

## What changes future outputs

Durable context changes how Superscale behaves beyond the current chat. Use it when you find yourself saying "always", "never", or "for this brand/product from now on."

<CardGroup cols={2}>
  <Card title="Save product facts">
    Put exact product details, claims, UI screenshots, packaging, and offer rules at product level.
  </Card>

  <Card title="Save brand rules">
    Put voice, logo behavior, colors, claim boundaries, ICPs, and visual no-gos at brand level.
  </Card>

  <Card title="Save workspace standards">
    Put agency-wide process, naming, approval, QA, and legal review rules at workspace level.
  </Card>

  <Card title="Save learnings">
    Turn strong reports, winning patterns, and repeated corrections into context so the next batch starts smarter.
  </Card>
</CardGroup>

## What stays temporary

Some information should not become durable. Keep it in the prompt or current chat when it only applies once.

* A weekend sale or short-lived promotion.
* A rough direction you are exploring but have not approved.
* A reference you want to imitate once, not forever.
* A competitor example that should inspire one batch only.
* A temporary constraint for a specific channel, market, or test.

<Tip>
  When in doubt, ask: "Should this affect future work?" If yes, save it as context. If no, keep it in the current task.
</Tip>

## Before you generate

Use this check when product accuracy, brand fit, or credits matter.

<Steps>
  <Step title="Choose the right brand and product" icon="folder-tree">
    Make sure the task is attached to the brand, product, collection, app, service, or landing page the agent should reason about.
  </Step>

  <Step title="Attach the current references" icon="paperclip">
    Add the exact product shot, app screenshot, logo, script, video clip, or winning ad that should guide this output.
  </Step>

  <Step title="Say what must stay fixed" icon="lock">
    Spell out the non-negotiables: product shape, packaging, UI, logo, claim, price, layout, voice, or disclaimer.
  </Step>

  <Step title="Decide what should persist" icon="refresh-cw">
    If the correction should apply again, save it to product, brand, or workspace context instead of leaving it buried in chat.
  </Step>
</Steps>

## Turning a correction into context

If an output is wrong, do not only fix the one asset. Decide whether the correction should become a rule.

| If the issue is                                               | Save it where     |
| ------------------------------------------------------------- | ----------------- |
| Wrong logo, color, voice, claim boundary, or brand tone       | Brand context     |
| Wrong product image, packaging, UI, feature, offer, or price  | Product context   |
| Wrong agency QA process, naming, review, or approval behavior | Workspace context |
| One bad prompt or one-off campaign instruction                | Current chat only |

<CardGroup cols={2}>
  <Card title="Levels & sources" href="/context/levels-and-sources">
    Decide whether context belongs at workspace, brand, product, or task level.
  </Card>

  <Card title="Reference images & winning ads" href="/context/reference-images">
    Use visual references when exact look, layout, logo, product, or style matters.
  </Card>

  <Card title="Asset Drive" href="/features/asset-drive">
    Understand the difference between storing assets and actively using them.
  </Card>

  <Card title="Keeping context fresh" href="/context/managing">
    Maintain context as products, campaigns, and learnings change.
  </Card>
</CardGroup>