Workstation runner
Runner overview
Status of this workstation runner: its link to the Skiller web app and the workstation Codex.
Skiller web connection
Checking…AI agents on this workstation
Checking…
Detected AI CLIs on this PC. Codex is the default for App designer and job execution; Claude, Gemini, and
Grok Build can run jobs when an app sets executionProvider (and optionally a command template).
Runner identity
Read-only snapshot. Use Settings to edit machine label, Skiller web URL, and pairing.
App packaging
App designer
Load a ready Skiller app bundle, or scan a project on this PC and let the workstation agent design one. Review the result locally, then push it to Skiller web as a private draft.
Use an existing bundle
Fast path
Select a prepared .skiller-bundle.json file. The Runner validates it locally and opens the
normal review and push controls without repeating project scanning or app design.
Project on this PC
Step 1
Choose the production project folder on this PC. The runner reads its optional
.skiller/project.json, then scans the declared skills directory (default skills/)
for folders containing SKILL.md and suggests app names and descriptions in the next steps.
Skills found
Step 2Choose the skills this app will use and set the order Codex should orchestrate them. Confirm before the app description step opens.
| Use | Order | Skill | Runtime role | Path |
|---|
Describe the app
Step 3After you confirm skills, the next available agent (by Settings priority) reads the selected skill instructions and drafts this description. Edit anything before designing the app — your edits are never overwritten.
Loading skill instructions…
Workspace user fields
Optionally add a field that defaults to the connected Skiller user or lets the end user choose an active workspace member.
Design and review the bundle
Step 4The next available agent (by Settings priority) designs the forms, files, skill workflow, prompt, and outputs. The result stays a private draft until an administrator reviews and activates it in Skiller web.
The basic template is a fallback and does not use agent intelligence.
A bird's-eye view of the app's purpose, workflow, and results. This optional image is attached when the draft is pushed.
View full bundle JSON
Skill repositories
Application skills
Local credential vault
Secrets
Store credentials on this workstation and approve each alias for an exact application configuration. Values never go to Skiller web or prompts; only approved local jobs receive them in their isolated environment.
Runner secrets
Checking…Values stay encrypted on this workstation. Skiller web receives only each alias, its app-scoped readiness, and no value.
Security boundary: all apps in one Runner profile share one trusted Windows OS-user zone. Use separate, restricted Windows accounts and Runner profiles for mutually untrusted apps.
| Alias | Value | Approved apps | Updated | Actions |
|---|
On this Runner
Apps
Assigned apps, their local projects, and the skill versions available to this Runner.
Imported app configuration
These local configurations are saved here even while another Runner is assigned. Choose Any connected runner in Skiller web to let prepared Runners share an app.
| App | Production project | Local setup | Actions |
|---|
Configuration
Settings
Edit pairing and machine label, then review what Skiller web and this workstation report.
This workstation
Checking…Edit what this runner controls locally. The connection chip shows whether Skiller web accepts heartbeats from this workstation.
This machine
Checking…Runner deployment and detected AI CLIs on this workstation — split into general runner info and agent capabilities.
General info
How this runner is deployed and where it stores job work.
AI agents on this machine
Detected AI CLIs. Drag cards to set priority for Step 3 analysis and Step 4 bundle design (exhausted agents are skipped). Gear opens default model and path settings. App designer packages apps; Run jobs can execute requests.
Step 3 analysis and Step 4 bundle design try agents in card order when compute is left. Ollama is a local model server card (not in that priority list) — open its gear for URL and Docker setup.

