Skip to main content

​
Three ways agents get information

Agents only know what you give them each turn. In XSpada, information reaches an agent through three distinct paths — and confusing them is one of the most common setup mistakes.

​
Linked context: permanent reference material

Linked context files are attached to the agent itself on the Linked Context tab. Their contents are included in the agent’s prompt stack whenever that agent runs — in Chat, the Lab, Actions, or Tracks. Use linked context for material that should always be true for this agent:
  • Policy documents and SOPs
  • Product catalogs or pricing sheets
  • Style guides and brand voice examples
  • Standard templates the agent should follow
Screenshot placeholder Agent Linked Context tab with two or three linked files listed. Show Link File button and a card per file with remove option.
Linked context counts against the agent’s context budget. Link only what this specialist needs — not your entire file library.

​
Chat-linked files: for this conversation only

During Chat or Lab, you can attach files to a single conversation — PDFs, spreadsheets, images, and other supported types. The agent can read and reason about them for that thread only. Use chat attachments for one-off work:
  • “Review this contract draft”
  • “Summarize this meeting recording transcript”
  • “Compare these two proposals I’m attaching now”
Screenshot placeholder Chat composer with attach menu open and one file chip visible above the input. Show linked file tile in the conversation area after send.
If the same file should inform every conversation with this agent, move it to linked context instead of re-attaching each time.

​
Artifacts: work products the agent creates

When an agent has the right builder Focus, it can create artifacts — structured outputs that open in a specialized editor beside the chat. Unlike chat text, artifacts are saved, versioned, and browsable in the Artifacts library. Artifacts are how agents produce durable work: documents, tables, components, code, images, and more. Each save creates a version you can revisit later.

​
Artifact types

During chat, new artifacts typically appear in a side panel next to the conversation. You can expand them, edit them, and continue instructing the agent to refine the output.
Screenshot placeholder Chat with split layout: conversation on the left, Text editor or Table editor artifact open on the right. Show agent message referencing the artifact it created.

​
How to ask an agent to create an artifact

  1. Assign the matching builder Focus — an agent without Table editor Focus cannot build a spreadsheet, no matter how you ask.
  2. Be explicit in your request — “Create a table comparing these three options” or “Draft this as a document I can edit.”
  3. Iterate in the same chat — artifacts update across versions; keep refining in conversation.

​
Need a builder type you do not see?

New artifact and builder types are added as XSpada expands. If your use case needs a capability that is not listed:
  • Describe the workflow to your XSpada rep — what the agent should produce and how your team would use it.
  • Use XSpada Evolve to explore whether an existing Focus combination already covers the job.
  • In the meantime, Text editor and Table editor cover many drafting and data tasks.
See Working with artifacts for the Artifacts library, version history, and share links.

​
Context engineering in practice

Remember the discipline from What agents are — and aren’t: relevance beats volume. Wrong context is as dangerous as no context — the agent sounds equally confident either way.

​
Where to go next

Working with artifacts

Browse, share, and review saved agent outputs.

Focus & capabilities

Match builder Focuses to the artifacts you need.
If you would like something unique to your business or have a unique issue, use XSpada Evolve or ask a XSpada rep.