Personalization

Prompt variables

Prompt variables generate text for an individual Page using saved instructions and selected prospect information. A template references the result with prompt.<slug>.

Documentation

Define the prompt

  1. Create a prompt variable with a clear name and instructions.
  2. Select at least one supported LinkedIn section.
  3. Save the definition and copy its exact reference.
  4. Use prompt preview to explore a sample result.
  5. Refine the instructions to shape the tone and style you want.

Supported section keys are full_profile, headline, job_title, company_name, experiences, and education. Choose only the context needed for the task. Saving a definition does not itself call AI.

<p>{{ prompt.opening }}</p>

The saved slug and prompt-variable name stay fixed after creation. Edit the instructions when the task changes; create a new definition when you need a distinct identity.

Understand execution

The Page owner must have the required definition. A shared template does not borrow the author’s prompt. Missing definitions, unusable selected inputs, and prompt execution failures can fail the PageRun.

Prompt instructions are plain instruction text; Liquid tags typed inside them are not an additional template-expansion system. The generated result is HTML-escaped when rendered and is not evaluated again as Liquid.

Preview and historical results

Generic template and extension previews can show blank prompt output. Rendering a Page does not initiate a new AI call. Changing instructions does not regenerate results already saved for old Pages. Inspect the generation result and failure reason rather than repeatedly refreshing the public URL.

Search guides and MCP references in one place.