Agents and executions

Overview

Agents and executions provides real-time visibility into how your AI agents process tasks and run technical skills on mobile. It allows you to follow multi-step reasoning workflows, observe tool invocations, and inspect executed code directly in your chat timeline.

Use this feature to verify agent behavior step-by-step, audit data calculations, and review tool outputs directly from your smartphone.

Prerequisites

To track agent and skill executions, you need:

• An active Swiftask account on the mobile app.

• An active conversation with an agent configured with multi-step workflows or technical skills (such as Python code execution).

Step-by-step guide

1. Track agent execution in real time

1. Select an agent or chat model from the Catalog tab under Recent, Chat, or My agents.

2. Enter your prompt in the Ask a question input field and send your request.

3. Notice the live execution block displayed beneath your prompt as the agent processes your request. Tap the block or the chevron to expand the reasoning stages and follow each processing phase.

2. Inspect skill execution and outputs

1. When an agent triggers an external tool (such as Python code execution or data visualization), a dedicated skill card appears in the chat timeline indicating the tool name and total steps (e.g., Python: Code + Upload + Visualization — 2 steps).

2. Tap the chevron (⌄) on the skill card to unfold the detailed execution steps, script logic, and input parameters.

3. Review the generated code, execution status, and produced outputs or visual elements directly within the expanded card.

Practical use cases

• Code execution auditing: Inspect Python scripts and data processing steps generated by an agent during mobile data analysis.

• Workflow verification: Check tool invocations, intermediate calculations, and execution statuses before accepting results.

Tips & best practices

• Collapse steps for scannability: Tap the chevron on completed execution cards to collapse code details and keep your chat timeline compact.

• Review intermediate thinking: Expand earlier execution cards if an agent regenerates an answer to see why adjustments were made.

Troubleshooting

Skill card details do not expand

• Cause: The skill is still actively processing or network connectivity was temporarily interrupted.

• Fix: Wait for the execution status to finish or refresh your conversation by returning to the chat list.

Additional resources

• Conversations — Interacting with agents and models in Chat on mobile.

• Prompt library — Browse and apply saved templates in mobile conversations.