[williamloo-scale]

agent-env env website put takes a site's server image from --server-dockerfile (stored with its build context, as an a2a agent's image is), --server-image-ref or GitHub, with --health-path. The older --backend-* flags still work, and a separate frontend image is optional and warned of. A new website_state_verifier step reads each checked session of a website env's sites, or a multi env's, at the site's /go?sid= through /svc/<name>/, the verifier endpoint CUA-Gym sites serve, and checks one dot path of its current_state, initial_state or state_diff each, with rows scored like the other verifiers. With env_version pinned to the version the task deploys, put and preflight refuse checks of sites that version doesn't serve.

Validation: unit suite passes, with tests for the CLI's three image sources and its refusals, check evaluation, pinned-version preflight, and execute reading two sites' /go through a stubbed gateway.

🤖 Generated with Claude Code

Confidence Score: 5/5

The PR appears safe to merge; no actionable new issues remain.

Summary

Adds server-image sources to the website CLI and introduces website_state_verifier.

• Website setup accepts one server image and makes the separate frontend optional.
• Tasks can check the state reported by each website site and session.

Diagram

sequenceDiagram
    participant Step as Website verifier
    participant Store as Env store
    participant Gateway
    participant Site
    Step->>Store: Read deployed env version
    Step->>Gateway: "GET /svc/name/go?sid=session"
    Gateway->>Site: "GET /go?sid=session"
    Site-->>Step: Session state
    Step->>Step: Evaluate checks and save score

Loading

_{Reviews (9) · Last reviewed commit: "feat(website): build a site's server ima..." · Reviewed by Greptile}