Agent assets

Provide approved links, text snippets, and images for agent outputs

How Agent assets works

Agent assets is a content library tool that gives your agent pre-approved material to reference while generating messages. You can store links, reusable text snippets, and uploaded images. This keeps every customer-facing output on brand and compliant — the agent selects from your curated library instead of generating content from scratch. Assets are matched by title and description, so well-named assets are surfaced reliably when the agent’s prompt references them.

Configuration

Inputs

Prompt · string

Instructions for when the agent should use the assets.

Assets · object · required

List of curated assets. Each asset has a kind, title, and description with kind-specific content fields.

Asset kind · enum · required

Options: link, text, image.

Title · string · required

Asset identifier. Use stable names so prompts can reference them consistently.

Description · string

Guidance for when to use the asset.

Link URL · string

For link assets, the destination URL.

Text body · string

For text assets, the reusable copy block.

Image URL · string

For image assets, uploaded image URL.

Outputs

assets · object[] · required

The full set of configured assets available to the agent during the run.

assets[].kind · enum · required

The type of asset: link, text, or image.

assets[].title · string · required

The title identifier used to reference this asset in prompts and agent reasoning.

assets[].description · string

The usage guidance that tells the agent when this asset is appropriate.

assets[].content · string · required

The resolved content — a URL for link/image assets, or the text body for text assets.

Use Cases

Brand-Safe Imagery

Condition: Agent is composing a lifecycle or marketing email
Action: Agent selects from uploaded, approved product images instead of generating or linking external images
Outcome: Every email uses on-brand visuals that have passed design review

Legal/Compliance Copy

Condition: Message content touches pricing, terms, or regulatory topics
Action: Agent inserts the matching disclaimer or legal text snippet verbatim
Outcome: Compliance language is consistent across all automated communications

Resource Links

Condition: Agent recommends a help article, doc page, or onboarding guide
Action: Agent pulls the canonical URL from the asset library rather than generating a link
Outcome: Users always receive correct, up-to-date links — no broken or hallucinated URLs

Product Documentation

Condition: Agent responds to a feature question or onboarding milestone
Action: Agent references stored text snippets describing product capabilities and setup steps
Outcome: Responses are accurate and aligned with the latest product documentation

Best Practices

Asset Organization

• Keep asset titles short, stable, and descriptive so prompts can reference them reliably.
• Use a consistent naming convention (e.g., legal-disclaimer-pricing, img-hero-product) across your library.
• Group related assets with similar title prefixes to make the library easy to browse and maintain.
• Add one asset per concept — avoid bundling multiple links or snippets into a single asset.

Content Governance

• Store only approved, reviewed content for customer-facing communication.
• Use the description field to specify exactly when each asset should be used so the agent picks the right one.
• Periodically review and remove outdated assets to prevent the agent from referencing stale content.
• Version sensitive copy (legal disclaimers, pricing) and update assets promptly when terms change.

Prompt Integration

• Reference asset titles explicitly in your agent prompt (e.g., “Use the legal-disclaimer-pricing asset when discussing costs”).
• Instruct the agent to prefer assets over generated content for any compliance-sensitive output.
• Combine image assets with text assets when emails need both visuals and copy from the approved library.
• Test that the agent selects the correct asset by running sample events through the agent preview.