dsh-plugin-wepre

AI generations can take tens of seconds or longer, and the loading gap often feels like dead time. WePre Next fills that gap by pushing a single-viewport, offline, sandboxed content card to the user while they wait. The catch is that the publishing workflow itself is fiddly: log in, upload, review the QA report, fix the card, re-upload, and repeat. dsh-plugin-wepre collapses that entire loop into one dsh conversation so the agent can write the card, ship it, read the QA feedback, patch it, and republish—all without a human in the loop.

The plugin exposes five tools covering authentication (SMS or email one-time code with a locally persisted session), card publishing (.html or .zip through the server-side QA gate), and QA report retrieval. The standout design is the automated QA fix loop: when the gate returns QA_GATE_FAILED, the agent reads the structured per-viewport issue list, modifies the card, and republishes under the same contentId. Before every upload it also runs a local pre-flight scan that flags banned tokens like fetch, localStorage, and Worker, plus external or root-absolute URLs, reducing wasted round-trips. Credentials resolve from a config token, an environment variable, or the stored session, so both interactive sessions and headless CI pipelines work out of the box.

If your dsh workflow needs to deliver single-screen content cards to WePre Next, or you are building an automated pipeline for "AI-waiting-period content," this plugin turns publishing into a routine tool call for the agent rather than a manual ops step.