the-launch-watch

Installation
SKILL.md

The Launch Watch

New products live or die on early signal, and watching each one by hand is exactly the chore that gets skipped in a launch week. This loop runs for a fixed window - typically 14 days - then stops itself. The bounded window is the point: an unbounded launch watch becomes a second daily report nobody reads.

Loop discipline. Read references/loop-cadence-guide.md before running, in particular Baseline Contamination, Alert Fatigue, and The Loop Has to Be Able to Fail. Its early-signal thresholds are fixed for a bounded window, so the fatigue budget and the failability rule matter more here than baseline drift: confirm the gate can actually fail and stop the spend.

How to run

  1. Which products launched, and on what date. The window is per product, not per store, so a launch on day 9 of another product's window starts its own.
  2. The window length in days, and the hard test budget for the window. Both are required. A launch watch without a spend ceiling is how a flop outspends a bestseller.
  3. Daily performance per product: impressions, clicks, sessions, add-to-carts, orders, revenue, and spend.
  4. The early-signal thresholds: minimum sessions before any judgment, plus the CTR and conversion-rate bars to clear. Ask for them. Thresholds copied from another store's launch are worse than no thresholds.
  5. The pre-launch readiness output, if the-launch-readiness-check was run, so a weak early result can be checked against a gap that was already known and accepted.
  6. The ledger, for the window's start date, spend to date, and prior runs in this window.

Method

Installs
2
GitHub Stars
9
First Seen
Aug 7, 2026