Replicating WorkBuddy's User Entry in DeepSeek Harness: A Slot Takeover Plugin
Any plugin that needs to replace a built-in DSH UI element hits the same single-slot priority shadowing and single-file client constraints. The pattern of keeping host data separate from client rendering also gives a clean upgrade path when a real backend becomes available.
The DSH plugin ecosystem uses a slot system where components register into named slots. To replicate WorkBuddy's user info entry, the author chose to take over the official settings.launcher slot rather than add a new row to sidebar.footer.action. That slot is single-typed and renders only the lowest-priority registration, so the plugin must register with a negative priority like -100 to shadow the official AccountMenu. The underlying account service keeps running, so settings, theme, and sign-out remain real host calls.
The plugin is split into host and client halves. The host runs in the Node process and serves a single GET /workbuddy-user/api/profile endpoint that reads profile.json for demo data like tier and points. The client is a 541-line single file because the browser module loader can only resolve host-shared modules, not relative paths. Real avatar and nickname come from ctx.remote.account.getProfile(), while the menu mixes those with disabled demo buttons for WorkBuddy-only features.
Three pitfalls dominate the debugging: CSS defined but never injected via a <style> element, slot priority conflict at 0, and an inject list missing remote.account, which silently prevents activation. The article insists on browser verification with a Puppeteer probe because curl and DOM checks cannot catch these failures. A full reproduction checklist runs from file creation through syntax checks, plugin install, and interface tests.
Taking over a single slot beats adding a list-slot entry because it replaces the official UI while preserving the underlying account service; this is a cleaner pattern for any plugin that wants to swap an entry point.
The single-file client constraint is a direct consequence of DSH's module loader design, and it forces plugin authors to inline everything, making large plugins harder to maintain but also simpler to deploy.
The article's emphasis on browser-level verification stems from the fact that plugin registration can succeed while the UI is broken; interface tests alone cannot see whether styles are injected or whether the plugin actually activated.
Separating host data provisioning from client rendering is a pragmatic way to keep a demo-data UI switchable to a real backend later, but it adds an HTTP round-trip that pure client-side plugins would avoid.