Context-aware Orylo Actions
Make reusable actions appear in the right context, combine the inputs they need, and choose exactly how they run and deliver results.
Sections in this article
Actions for your current context
An Orylo Action combines three decisions: when it is available, what it does, and where its result goes. The Action Bar matches your current context to each enabled Action’s rules; you do not need to copy everything into a new conversation first.
Context-aware means matching captured context to your configured rules—not continuously watching the screen or automatically running an Action. You choose which matching Action to run.
| Context | What it contributes |
|---|---|
| Selected text | The text you selected; useful for rewriting, translating, or Promptize. |
| Files, folders, or images | Selected files or images, or folder paths, according to the configured input. |
| Web page | The current page URL—not an automatic copy of the page body. |
| Focused window | App identity, the available window title, and an on-demand screenshot; capture requires permission. |
| Application | The active app identity, including an optional specific-app condition. |
Decide when an Action appears
Choose a Primary Context: it is required and defines the main input. Add All Of conditions when every additional context must be present, or Any Of conditions when at least one alternative must be present. AI Actions receive all configured contexts that are actually available, not just the first matching alternative.
For example, require Selected Text plus Web Page to offer an Action only when text and a page URL are available together. To restrict an Action to a particular app, make Application a required condition and choose that app; an optional Application condition alone does not guarantee exclusivity.
These rules control availability. They do not grant file access or screen-recording permission. If an Action is missing, check its enabled state, required contexts, source app, and capture permissions.
Choose the operation and result
For AI results, choose the result panel, clipboard, Composer, replacement of the original text selection, or no automatic delivery. Background command output can go to the panel, clipboard, or Composer; opening a new Terminal does not collect output into Orylo. Opening a web page simply opens the configured URL.
Customize the Action’s name, description, icon, enabled state, and ordering as well as its behavior. Review commands before running them: they execute on your Mac and are not made safe merely by being an Action.
| Operation | Customization and boundaries |
|---|---|
| AI Action | Choose an Assistant and instructions. It receives the primary and available configured extra contexts. Chat Assistants generate from supplied content; Agent Assistants may use their configured tools. |
| Open Web Page | Set a URL template. Selected text or a page URL can fill its encoded context placeholder. Extra conditions only decide availability; they are not passed into the URL. |
| Run Command | Set the script, working directory, timeout, and background or new Terminal execution. Only the primary input is passed as data; app/window inputs currently provide no context text to the script. |
Create with Chat—or configure it yourself
Describe the workflow
Open Settings → Actions → add → Create with Chat. Specify the primary input, required context, operation, Assistant where needed, and delivery. You need an already working Agent Assistant for creation.
Check the proposed definition
Ask for a preview and confirm the matching conditions, instructions or command, target Assistant, and output destination. Save only the intended configuration; saving does not execute the Action.
Try it on a small example
Return to the intended app and context, open the Action Bar, and run the saved Action deliberately. Refine it in Settings or continue the creation conversation. Create Manually is also available.
Organize choices with Action Groups
A Group is a menu of alternatives, not a sequence that runs every child. For example, a selected-text group can offer an AI explanation, a web search, and a local command. You open the group and choose one Action.
Children share the group’s input and appearance conditions, but may use different operations and result destinations. Groups cannot be nested. An entirely disabled group has no usable entry.
Run and follow an execution
Bring the context
Select the text, files, folder, or image you need, or focus the relevant app/window. Open the Action Bar with the shortcut configured in Settings.
Choose a matching Action
Inspect its intent before running. An AI Action uses a separate execution session, not your open conversation’s history. Do not assume it carries the context of a previous chat.
Inspect progress and output
Open Action Executions from the expanded Action Bar to view running and recent results, partial output, and errors. Hiding the bar does not cancel a run; explicitly stop it in the execution detail.
Keep control of delivery and permissions
Automatic replacement is only available for an AI Action with selected text as its primary input. It depends on the source app and the original selection still being valid and editable. If delivery fails, inspect the retained result rather than rerunning a potentially side-effecting task.
Actions do not offer an interactive permission-approval flow. Requests requiring a new approval are not silently allowed; use a normal conversation for work that needs that interaction. Existing Agent permissions still matter.
Execution records are temporary for the current app session, not permanent conversation history, and do not survive an app restart. Copy or save results you need to keep. Optional Action suggestions are off by default and produce reviewable drafts, not automatically enabled or executed Actions.