launch-monitor

Installation
SKILL.md

Applies from the moment a launch is public until the coverage wave dies — typically two to four weeks. Produces a triaged response war room report where every signal carries an urgency level, an action badge, and an exact source link, so the team knows what to respond to, correct, amplify, or escalate — in that order.

Replace before enabling: {{ALERTS_CHANNEL}} — where act-now items get flagged to the team; {{COMMS_OWNER}} — who receives escalations (comms, legal, or leadership).

Before the first sweep

Do two research steps before asking the user anything:

  1. Resolve alternate names silently. Search the live web for the product's codenames, former names, version names, API/SDK/repo names, and model identifiers — they are often different from the marketing name, and coverage splits across all of them. Fold every confirmed variant into your query list without showing the user this step. Query patterns are in references/sources.md.
  2. Confirm the launch date from evidence. Search for when the product actually launched. Present what you found for confirmation; if announcement date and general availability differ, surface both and ask which window matters.

Then confirm in a single message: launch date, lookback window (default: launch to now), depth (quick scan vs deep sweep — default deep), and competitors to prioritize or exclude (default: on, identified automatically). If the user says "just run it", take every default. If the product name is ambiguous after research, disambiguate before spending a single search on the wrong product.

Profile the product before querying. Answer five questions from the announcement and a quick search: What category is it? What ecosystem does it live in? Who is the audience? What are the key launch claims? And — most important — what would a mischaracterization look like: wrong category, wrong price, wrong capability, wrong comparison? This profile decides which communities you sweep, and you cannot flag a mischaracterization without knowing what "correct" looks like.

The sweep

Run parallel searches across four fronts — press, community, social, competitors — highest-reach sources first, because that is where a wrong claim does the most damage before you catch it. The full tiered source list, per-platform query patterns, and consumer-launch additions are in references/sources.md; read it before the first sweep of any launch.

Installs
2
GitHub Stars
179
First Seen
10 days ago