When it comes to working with agents, context is crucial. In mabl, you can now apply custom agent instructions that match your team’s testing standards and application requirements. Agent instructions are user-defined rules that act as a context-aware extension of your team’s testing strategy. Use them to embed business logic, application-specific quirks, and governance rules directly into the context that mabl carries into every agentic task.
Agent instructions apply to the following capabilities:
- Authoring: explain how to interact with your app and create steps that follow your established conventions.
- Runtime recovery: give the agent more detailed guidance on how to respond when a step fails at runtime
- Results analysis: give the agent more detailed guidance on how to investigate failures and summarize run outcomes
Try it out
To add an instruction, go to Agents > Settings. Click on Add instructions.
Provide a short piece of natural language guidance up to 1,000 characters that an agent considers while carrying out its task. Examples of instructions could be:
- “When creating variables, use the naming convention prefix_descriptive_name.”
- “For the checkout application, always verify the total price after adding items to the cart.”
- “For the checkout application, treat any failure on the /payment endpoint as high priority and surface it at the top of the summary.”
Optionally add one or more of the following scopes:
- Capability: authoring or runtime recovery
- Application: one or more applications
- Environment: one or more environments
If you don’t apply a scope, the instruction applies at the workspace level.
Use the on/off toggle on the agent settings page to disable an instruction without deleting it. Disabled instructions are saved but not sent to agents.
Share feedback
If you have any feedback on agent instructions, we highly encourage you to share it with your mabl customer success manager (CSM). Your insights help us build a better product!
Learn more
Learn more about best practices for writing agent instructions in the docs.