@xevy/skills 0.1.40 → 0.1.41
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +1 -1
- package/package.json +1 -1
- package/skills/computer/SKILL.md +10 -2
package/README.md
CHANGED
|
@@ -137,7 +137,7 @@ Use generate-guide to make docs/guides/customer-form/manual.md with screenshots
|
|
|
137
137
|
|
|
138
138
|
Computer works across desktop and browser apps and projects. Windows uses the signed-in desktop; macOS 14+ requires native Screen Recording and Accessibility consent. The guide workflow stays in the active engine, respects blocks and human take-over, and identifies incomplete verification or publication. Execute/fix use it for UI journeys when an authorized test environment is available.
|
|
139
139
|
|
|
140
|
-
The
|
|
140
|
+
The package version is `0.1.41`. Run `npx @xevy/skills update -y` after that version is published.
|
|
141
141
|
|
|
142
142
|
## Security
|
|
143
143
|
|
package/package.json
CHANGED
package/skills/computer/SKILL.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: computer
|
|
3
|
-
description: Operate connected Windows or macOS applications through Xevy Computer,
|
|
3
|
+
description: Operate connected Windows or macOS applications through Xevy Computer, including selected browser profiles and tabs. Use for desktop or browser tasks across apps and projects, including walkthroughs, testing, and user guides.
|
|
4
4
|
---
|
|
5
5
|
|
|
6
6
|
# Computer
|
|
@@ -15,6 +15,14 @@ Start an exclusive `agent` session with `computer_start`, associating the curren
|
|
|
15
15
|
|
|
16
16
|
Inspect an existing controller before changing it. Human take-over advances the generation and owns control. Continue only when control is explicitly returned; refresh session metadata before further input. Send the captured generation with pause/resume/stop, including cleanup, so an old agent cannot stop a new controller. Stop only the session this run owns. For owned cleanup, call `computer_control({computerId, sessionId, generation, transition: 'stop'})` with the captured generation. On the public surface, start with `computer_start({computerId, mode: 'agent', projectId, taskId})` using relevant context; the internal server surface takes only `computerId` and inherits its task, so follow the active tool schema.
|
|
17
17
|
|
|
18
|
+
## Browser control
|
|
19
|
+
|
|
20
|
+
For browser work, use tab-scoped browser actions when the companion advertises them. Start an agent session, then call `browser.discover` through `computer_action` and read its `computer_command` result. Select the browser and profile named by the user; otherwise use the relevant open browser/profile shown by discovery. Do not infer a profile from a window title alone.
|
|
21
|
+
|
|
22
|
+
Call `browser.connect` with the selected `browserId` and `profileId` to attach to an existing local debugging endpoint. A running browser with a usable endpoint should remain open. If there is no endpoint, explain that starting debugging requires closing and reopening the selected browser/profile. Send `restart: true` only when that restart is authorized; never terminate another profile to obtain a port. The companion owns the loopback debugging connection. Do not connect directly to a debug port from the agent host or expose it across the network.
|
|
23
|
+
|
|
24
|
+
Use `browser.tabs` to select the actual tab, then `browser.navigate`, `browser.snapshot`, `browser.click`, `browser.type`, and `browser.key` as needed. Read every command result. Browser coordinates belong to the selected tab viewport; they are distinct from desktop screen coordinates. Prefer `browser.snapshot` for evidence because it captures the tab without unrelated desktop windows. Do not call `desktop.snapshot` as a substitute when a tab-scoped capture is available or a full-monitor capture is blocked. `browser.disconnect` releases the companion's browser connection; it closes a browser only when the companion owns that launch. Follow the tool's current schema and capability list, since older companion versions may offer desktop actions only.
|
|
25
|
+
|
|
18
26
|
## Observe, act, verify
|
|
19
27
|
|
|
20
28
|
1. Request `desktop.monitors` and `desktop.windows` through `computer_action`, then read their `computer_command` results. Select the actual target window and monitor.
|
|
@@ -22,7 +30,7 @@ Inspect an existing controller before changing it. Human take-over advances the
|
|
|
22
30
|
3. Queue bounded actions with the current generation: focus, launch a supported application, click, drag, scroll, type, or press a key. Use observations to choose the next dependent action.
|
|
23
31
|
4. Read each command until a terminal result, then take a fresh screenshot and check the expected app result. For a transaction, verify its final state, such as the saved file, submitted record, or visible deletion confirmation. Native input success establishes that the action executed; business success requires observed application evidence.
|
|
24
32
|
5. A timeout or lost acknowledgement leaves the result uncertain. Inspect command status and the desktop before retrying a write. Resume from the observed state; avoid repeating a submit, payment, send, or delete whose outcome is unresolved.
|
|
25
|
-
6. After all operation commands finish and a final screenshot is observed, call `computer_complete_operation` using the same `operationId`, current generation, and `completed`, `submitted`, `deleted`, or `failed`. Summarize the app result actually verified. Stop owned control when finished.
|
|
33
|
+
6. After all operation commands finish and a final desktop or browser-tab screenshot is observed, call `computer_complete_operation` using the same `operationId`, current generation, and `completed`, `submitted`, `deleted`, or `failed`. Summarize the app result actually verified. Stop owned control when finished.
|
|
26
34
|
|
|
27
35
|
Use the actual tool argument names and response envelope. Queue `const {command} = await computer_action({computerId, sessionId, generation, action: 'desktop.snapshot', payload: {}, operationId, summary: 'Verify the test record'})`, then read `await computer_command({computerId, commandId: command.id})`. The action response is `{command: {...}}`; take the ID from its nested `command`, while MCP clients parse the response's JSON text content. Read each action's current payload schema before constructing inputs.
|
|
28
36
|
|