The r2 spec review on #104 produced a complete green verdict and then failed
on --max-turns 40 at turn 43, so the label step never ran and the transition
had to be reconciled by hand. Forty was a guess; a review that reads SCOPE,
AGENTS, PROCESS, the issue thread and the spec exceeds it routinely, and a
code review that also runs gates needs far more.
The real guard against a runaway run is the job timeout, not the turn count.
Issue: #114
User-Visible: no
The first live run failed with "Could not fetch an OIDC token": the action
needs id-token: write to authenticate the GitHub App.
The reviewer also checked out dev, where the material under review does not
exist yet — specs and code are committed to issue/<NN>-slug. The job now
switches to that branch when it is pushed, and warns loudly when it is not.
Issue: #114
User-Visible: no
Adds .github/workflows/process.yml. A status label change is the trigger:
S4-spec-review runs the spec review, S7-code-review runs the code review,
and the verdict decides the next label. Only a green verdict advances;
yellow and red return the task to its author. Cycle limits (4, or 2 on the
light track) are counted from the verdicts already posted on the issue.
Labels are moved with HP_PROCESS_TOKEN, not GITHUB_TOKEN, so the change
emits an event and the chain continues.
Issue: #114
User-Visible: no
HACS will install from the named asset instead of the auto zipball, so
GitHub's public download counter becomes a per-version install metric.
The workflow also has a manual dispatch to backfill an existing release.
T1: demo/serve.mjs exports check/checkAll/finish — all 48 smokes now
assert named facts and exit non-zero on a mismatch or an uncaught
in-card exception (verified by breaking the kiosk guard on purpose).
Informational values were frozen from a v1.43.1 run and cross-read
against the source; timings assert budgets, not exact numbers.
T2: new CI job 'smoke' gated on 'frontend', builds a FRESH bundle
before running (the committed demo/srv/assets copy is a snapshot) and
uploads per-file logs on failure.
T3: [auto] now means 'a named failing check exists' and each line names
it (43 lines); 72 aspirational markers honestly downgraded to [manual].
Fixed the 'ZERO edit buttons' contradiction (wrong since v1.30.1) and
the opening-click line (true again since v1.43.1).
Three smokes carried pre-v1.39.0/v1.25 expectations and were testing
old behaviour: tap defaults for lights, card-wide tap action, label drag
requiring plan mode.
DEVELOPMENT.md documents the harness contract.
Matysh