reoring

herdr-event-watch

Event-driven watching of herdr lanes, durable inbox artifacts, and PR required checks, instead of fixed-interval polling. Use when supervising worker lanes ("tell me when X finishes"), waiting on PR checks before a merge decision, or when periodic patrol keeps reporting "no change". Runs scripts/herdr-event-watch.sh in a background terminal and reacts to emitted event lines.

main-po-patrol

The main PO's patrol over the fleet's project POs — check each PO session's health (context usage, degradation symptoms, forced model switches) and drive session rotation via /po-handover → /po-resume where needed. The subjects are POs, not lanes. Triggers - "main PO patrol", "check on the POs", "PO health sweep", "PO context check".

po-handover

Write a PO-session handover file and rotate the session safely — the PO counterpart of disposable worker sessions. Run at 50-60% context usage, or at a natural boundary (a gate just closed, a wave just finished), BEFORE judgment degrades. The new session resumes with /po-resume. Triggers - "po-handover", "prepare the handover", "rotate the PO session".

po-resume

Resume a PO session from a /po-handover file. Read the handover as a map, rebuild state from live measurement, re-arm monitors, and load the standard PO toolkit (dev-flow / issue-lane / herdr-event-watch plus your fleet's ledger and messaging conventions) into working awareness before doing any PO work. Triggers - "po-resume", "resume the PO session", "take over as PO". Argument - path to the handover file.

watch-po

Event-driven detection of stalled or abnormal PO panes, reported to the main PO — approval-prompt stalls (POPROMPT), context threshold crossings (POCTX), forced model switches (POMODEL), pane disappearance (PODEAD) — via a persistent Monitor, optionally pushed to a notification topic. The standing sensor that fills the gaps between main-po-patrol rounds. Triggers - "arm the PO watch", "watch-po", "tell me when a PO stalls".

dev-flow

Lightweight development flow — the default track for everyday work. source-only changes, dev/staging live applies, and zero-user production (owner-declared). Evidence is a merged PR + green CI + one ledger line; no sealing, no successor tickets, no verdict ping-pong. Routes irreversible or customer-visible work to prod-flow via a one-question track test. Keeps hollow-green defenses and a fix-plus-regression-gate rule for problems found live.

prod-flow

Full-rigor staged flow for irreversible or customer-visible work — concept → read-only scout → design → PoC → certify → implement → manual live apply → observe → confirm. Use when a failure would consume a non-refundable consumable (approval, signature, publication, quota) or be visible to a real customer. Enforces packet sealing (GO bound to a packet digest), per-stage receipts, fail-closed gates, and review round limits.