Skills#
Skills are small non-personalized RPA robots developed in Sherpa Designer that always perform one specific operation (for example, generating/processing a document).
They imitate human actions and consist of 1–3 steps (without collecting primary data).
Skills eliminate errors due to templated execution, but remain inflexible (they are difficult to change quickly without the right specialist).
Skill Input and Output Arguments #
When developing Skills in Sherpa Designer, input and output arguments are provided (this simplifies integration into Sherpa Autopilot).
In chat mode, a Skill runs:
- With user confirmation
All input arguments are shown as fields prefilled by the neural network. The user can confirm launch, edit values, or decline. - Without confirmation
The scenario starts instantly. To activate, select the checkbox "Allowed to run without user confirmation of input parameters" in the robot properties.
Skill Interaction with the User in Chat#
Skills can communicate directly in the chat window. During interaction they can:
- display statuses and text,
- request input in a text field,
- offer a selection from a predefined list,
- require confirmation with buttons.
After the user sends a response, the scenario continues automatically.
Example: When searching for a person's contact in the address book, the Skill displays buttons with possible options (if multiple matches are found). The user can click the required button, and the scenario resumes with the selected contact.
Task Decomposition and Skill Chains#
The neural network analyzes a complex task, selects several Skills, and executes them sequentially.
The output data of the previous Sherpa Autopilot dialog is passed as input data for the next task.
Example: Task “compose and send an email to Vladimir Mikhailovich”:
- option 1: the “sending” Skill finds the email in the address book itself;
- option 2: a separate “Address Book” Skill returns the email, which is passed to the “sending” Skill.
Scriptless Robot Mode#
If there is no suitable Skill, but the task can be completed using desktop tools on the user's computer, the Scriptless Robot is activated.
The Scriptless Robot:
- requires clicking the target window;
- captures mouse/keyboard, clicks icons, and enters text in applications/browser;
- can be interrupted at any time and shows confirmation before dangerous actions.
The completed Scriptless Robot sequence can be:
- saved as a scenario and opened in Sherpa Designer for editing,
- published as a new Skill in two clicks.