scripts/verify-deploy.sh checks a pds.did.bot deploy against a baseline. before records the time, the served revision, the process's anonymous memory, and the kinds of warning the journal logged over the last 24 hours. after checks that /xrpc/_health reports the expected revision, that each restart's outage (journal no longer accepting → listening (tls)) is within budget, that no new warning kind appeared, and that memory is under a ceiling. --post-as <did> also writes one scrobble through that account's session and reads it back. It exits non-zero with one line per failed check.
scripts/ssm-run.py reads the instance over SSM with boto3 via uv, using the caller's own AWS profile. docs/deployment.md gains a short "Checking a deploy" section.
Dry run against production, with a baseline rewound to before yesterday's restart, then with tight thresholds:
🤖 Generated with Claude Code