surface

Skills.

Choose the workflows Penny can execute in Lab, configure required plugins, and create or import reusable Skills that produce canvas results.

Written for
Anyone using or creating Lab Skills.

A Skill gives Penny a reusable action in Lab: generate an image, transform a reference, produce audio, or run a workflow you have saved. Penny supplies the inputs, the Skill executes its underlying work, and supported results appear on the canvas.

Use Playbooks for written instructions and preferences. Use Skills for actions Penny can execute.

Choose available Skills

Open Lab, then click the Skills wand beside the conversation selector in the message bar. The count shows enabled Skills out of the total. The menu opens upward and groups Skills by their output: Image, Audio, Video, and Text.

Find the Skills control beside the conversation selector in Lab.

The Skills popover opens above the wand in Lab's message bar. Each row has a name, description, and enable/disable switch on the right. Skills are grouped by Image, Audio, Video, and Text outputs. Create Skill and Import Skill are at the bottom. The wand's count is enabled Skills / total Skills.

Enable the actions Penny should be able to use. This changes availability; it does not start work. Penny cannot call a disabled Skill, although it may still know that the Skill is configured and explain why it is unavailable.

An enabled Skill can also be unavailable when its required media is missing. For example, an image-editing Skill that requires a source image needs an image on the canvas. A description of an image does not replace that input.

Ask for a specific result

Ask for a concrete result and mention the canvas item that should be its input. Kaitoi resolves the item reference to its underlying media. Click the item's handle label or use @ in the composer to insert a mention. Include the transformation and useful constraints, such as dimensions or duration.

A Skill exposes useful inputs while keeping the workflow steps together.

Example: a custom Square social image Skill takes a canvas image and a target size. Penny resolves @reference to the actual image and supplies 1080 as the size. The saved workflow center-crops and resizes it, returning a square image to inspect on the canvas. The image and size are exposed inputs; the crop and resize operations remain inside the workflow. This is an illustrative Skill, not a built-in name or an executed result.

Enabling a Skill does not force Penny to call it on every message. Ask for the intended action and follow the particular execution's queue, progress, or error indication. Inspect the resulting media before requesting another variation.

Make a custom Skill

Choose Create Skill at the bottom of the browser, then pick a source:

  • From Node wraps a supported node. Use it when one operation already does the job.
  • From Template wraps a saved workflow. Use it for connected steps you want Penny to call together.

For a template, expose the pins Penny should supply or return when saving the template. The create dialog lists the inputs and outputs for review. If none were explicitly exposed, it can detect unconnected pins; review those choices before creating the Skill. Without usable outputs, Penny cannot retrieve the workflow's results.

Give the Skill a clear Name, Description, and Output type. Describe when to use it and what inputs it needs. For example: “Crop an image to a square and resize it for a social post; requires a source image.”

Some nodes need a plugin, such as an LLM plugin. The create dialog lists those inputs and asks you to choose the plugin. Fill every required plugin choice before creating the Skill. If none is offered, read the dialog's reason: a compatible plugin may be missing, or its required API key may be unavailable. Plugin bindings are part of the Skill's setup, rather than text Penny invents when calling it.

If the dialog says the node needs another node connected to a required pin, build that connection in Graph, save the workflow as a template, and use From Template instead.

Test the Skill on a small input. Confirm the output type, media format, and provider requirements before relying on it in a larger conversation.

Import, export, and repair

Choose Import Skill to load a .kaitoi-skill.json file. For a Skill you own, open its Skill actions menu to Edit, Export, or Delete it. Deletion asks for confirmation and cannot be undone.

A Skill file describes a workflow. Importing it does not create missing provider credentials or make an unavailable model accessible. Review embedded graph inputs and plugin settings before sharing an export. Test an imported Skill with the destination account's actual providers and capabilities.

An older Skill that asks Penny to supply a required plugin as plain text needs reconfiguration. Recreate your own Skill with an explicit plugin choice; a built-in or shared Skill may need its maintainer to update it.

Execution and cost

Skills execute the underlying work and inherit its costs and restrictions. Turning a workflow into a Skill does not make provider calls free. Check the result and execution status before asking for another variation.

When a Skill does not work

What you see What to check
Penny says the Skill is disabled Enable it in the Skills browser.
It is enabled but cannot be called Check the required media types on the canvas and mention the intended input.
The create dialog cannot satisfy a required connection Configure the plugin, or connect the required node in a template.
Work is queued or still running Follow that execution's status before starting another request.
An execution fails Read its error, check provider access, and inspect any items already produced. Retrying can repeat paid work.

See Working with Penny for stopping and recovery, Lab for arranging references, and Files and results for retaining media.

Last reviewed against Kaitoi on 24 October 2026. View this page as Markdown

Type at least two characters.