factory-report

Give a snapshot of where the factory pipeline stands, across Linear, right now. Read-only — this never claims, triages, or dispatches anything.

1. Resolve scope

Default: every repo in config/repos.yaml. $ARGUMENTS, if present, names one repo (matches factory queue --repo <name>'s matching).

2. Gather

From the factory checkout (or any cwd with factory on PATH):

factory queue --repo <repo>      # omit --repo for every configured repo
factory digest --repo <repo>     # omit --repo for every configured repo
factory ci-health --report       # base-branch CI alarms (add --repo to narrow)

queue gives per-repo counts (Triage, Todo-not-ready, READY to dispatch, In Progress, In Review, Blocked, Done/total) plus which tickets are startable right now and how many open PRs are waiting on merge. digest lists every ai:blocked hold with its age and whether a reply has landed (which means the next triage tick will re-examine it). Both are cheap, read-only, and already read the same queue snapshot the pipeline itself uses — don't re-derive counts by querying Linear directly.

ci-health --report reads only the alarm state factory ci-health last wrote (~/.factory/state/ci-health.json) — it makes no GitHub calls, so it is as cheap as the other two. A RED <job> line means that job has failed on that many consecutive pushes to the repo's base branch and the operator has already been notified once; a repo that has never been checked says so rather than claiming green.

3. Report

Synthesize, don't just relay raw terminal output (the ANSI coloring won't survive into chat). Structure:

• Headline — total tickets across scope, and the one-line pipeline constraint: is it starved for specification (deep Triage, empty ready queue), starved for dispatch slots (ready tickets but no free workers), or clear?
• Per-repo table — repo, Triage, Todo-not-ready, Ready, In Progress, In Review, Blocked, Done/total. Flag any repo where Triage is unusually deep (>20) or Ready sat nonzero with zero free slots.
• Ready to work now — the startable tickets queue printed, grouped by repo, so this doubles as "what can we work next."
• Needs a human — every held ticket from digest, oldest first, with its question excerpt; call out any marked ANSWERED since those are one triage tick away from unblocking, not stuck.
• Awaiting review/merge — repos with nonzero In Review or open PRs.
• Base CI health — one line per repo from ci-health --report. A red base blocks every merge and every ship for that repo, so say it here even when the queue looks healthy; name the job and how many pushes it has been red. "Never checked" is not green — report it as unknown.