Verify an agent-built app on its deployed URL with a harmless write, a read after reload, and an access-control check. Pair any failure with service logs and resource metrics from the same interval.
We are the team behind Lizard (lizard.build). Our October 8 announcement describes deployment through a terminal or GitHub, managed services, and live observability.
Keep the deployment route explicit. lizard up uploads local code; an existing GitHub-linked app should follow its repository deployment workflow. Record the deployed revision before diagnosing behavior.
For a database-backed app, explicitly provision Managed Postgres and connect the consuming service. The documented provisioning command is lizard add postgres; application configuration and migrations are separate work. A successful deployment does not demonstrate a working connection.
For a hypothetical issue tracker, create a disposable issue on the real URL, reload it, and check that another account cannot read a private issue. These are proposed acceptance steps, not claimed test results.
On failure, record the request time, expected behavior, and observed result without credentials or personal data. Inspect the corresponding service logs; compare metrics for signs of resource pressure. Ask the agent for a change justified by that evidence, then repeat the failed step. Keep a human responsible for secrets, authorization, and recovery decisions.
October 8 announcement · Lizard documentation · Explore Lizard
