rig/test
dan-claude-bot 34f1986da0 feat(bootstrap): prove box landed on PATH after a claimed install success — don't trust exit codes (#12)
Issue #12's review comment named the failure shape exactly: box's
setup-host is written for a sudo-capable user and one of its paths exits
0 after only adding a group, asking for a re-login — so an installer's
exit code can claim a success that never took effect. That is the sshd
first-wins bug's shape, and rig's doctrine is to assert effective state.

The check stays deliberately light: command -v box proves the one
artifact rig asked the installer for. Anything deeper — daemon, pool,
network — is box's domain; rig never interrogates Incus, so the success
log hands the operator 'box doctor' (box's own effective-state verdict)
instead of reimplementing it. A hollow success WARNS with the manual
pointer, never dies: box is the host extra, and the OS+tailnet core is
already done and asserted by the time this block runs.

Tests grep the shipped script (the check needs root + network to
exercise): the call, the warn wording, the delegation to box doctor,
and a fail-closed line-number assert that the check follows the
installer run. Rides along: the README rename greps (#12) — the stale
heavy-duty/claudebox slug is negative-grepped out for good.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-18 14:50:46 +00:00
..
cli.sh feat(bootstrap): prove box landed on PATH after a claimed install success — don't trust exit codes (#12) 2026-07-18 14:50:46 +00:00
db-integration.sh test(db): add real dump/restore round-trip probe + CI job + manual proof docs 2026-07-17 15:48:51 +00:00