gamebeastgg/gamebeast

Playbook for building a project's first or missing dashboards from the events it actually sends: engagement, retention, and (when the data exists) economy and monetization boards, each chart run and checked before it's saved. Load when the user wants dashboards built, or asks what they should be tracking on a board.

Full reference for cohorts: reading definition JSON and building new player segment definitions. Load to inspect/explain existing cohorts or understand the AST.

Playbook for investigating why a metric moved (or why events stopped): pin the change in time, localize it to a segment, check configuration edits and experiments around that date, and recommend a fix with evidence. Load when the user asks why something changed, dropped, spiked, or stopped.

Gradual (staged) rollouts for Gamebeast experiments: how the rollout gate and its smooth ramp curve work, the curve presets, turning a ramp request into rollout steps for create_experiment, reading a rollout from get_experiment and get_experiment_results, and operating one with update_experiment_rollout (set, advance, pause, resume, replaceSteps). Load before designing, changing, or explaining any experiment rollout.

Playbook for auditing and fixing event data: events that stopped, properties with drifting types, high-cardinality traps, undocumented or noisy events, and tracking-plan mismatches. Load when the user asks why an event looks wrong, whether their data is trustworthy, or to fix what a tracking plan flagged.

Full reference for ad-hoc funnel queries: steps, conversion window, measurement, step ordering, constant attributes, filters, breakdowns and breakdown limits, first-time/new-user filters. Load to build, read, or explain a funnel spec.

Full reference for heatmaps: the difference between a heatmap (the map definition on the Data page) and a heatmap query (a chart that counts an event's positions on that map), the spec, fetching the map image, saving heatmap queries, and how to read hotspots. Load to build, read, save, or explain anything heatmap.

Full reference for ad-hoc insight queries: spec format, metric types, aggregations, formulas, filters (incl. cohort, experiment, deeplink, array), breakdowns and breakdown limits, first-time/new-user filters, and per-metric value formatting (display.unit, compact, decimals). Load to build, read, or explain an insight spec, or before setting a metric's unit.

Playbook for instrumenting a game or app with Gamebeast: design a tracking plan of events and the tuning values worth moving into a remote configuration, hand the code changes to a coding agent (or make them, if you can edit the code), and verify each event against live data and each value against the config. Load when the user wants to set up or extend their analytics events, make their game tunable or experiment-ready, or asks what they should track.

Playbook for running a measurable LiveOps event in a game (a double-XP weekend, a limited-time shop, a seasonal currency): configuration changes shipped as a holdout experiment with a staged rollout, a KPI dashboard, and a readout afterwards. Load when the user wants to plan, launch, or read out an in-game event.

Create, edit, describe, and delete saved cohorts (player segments) with create_cohort, update_cohort, and delete_cohort. Load before any cohort write.

Create, edit, lay out, organize, copy, and delete analytics dashboards and saved queries: saving an ad-hoc chart, editing a saved query's definition, placing and sizing panels on the grid, text panels, folders, and what each write keeps or deletes. Load before any dashboard or saved-query change.

Describe analytics events (markers) and their arguments, and hide or show them in the query builder — including events the game does not send yet. Load before calling describe_event_types.

Create, terminate, and manage Gamebeast A/B experiments: how to design a definition (groups, weights, changesets, targeting, staged rollout), check it with preview_experiment, create it with create_experiment, end it with terminate_experiment, ramp it with update_experiment_rollout, and pin or exclude individual units with assign_experiment_unit. Load before making any experiment change.

Playbook for turning community feedback into a testable change: find the topic players care about most, check it against behavioral data, find the configuration lever, and recommend an experiment or fix. Load when the user asks what players are complaining about or how to act on feedback.

Full reference for ad-hoc retention queries: start and return events, retention window and group (day-N / week-N), look-back vs look-ahead, displays (series, curve, single number), filters, breakdowns, and how to read the results. Load to build, read, or explain a retention spec.

Execute saved analytics queries (insights, funnels, retention, heatmaps) with optional time/breakdown overrides, and interpret insight, funnel, retention, and heatmap results.

Playbook for deciding running experiments: read each one's results, decide ship, end, extend, or ramp with explicit rules (sample-ratio checks, power, duration), and apply the decision. Load when the user asks whether an experiment is done, which one won, or what to do with running experiments.

Playbook for finding the players a change should reach and reaching them: build and sanity-check a cohort definition, then target an experiment at it. Load when the user wants to reach a specific group of players, or asks which players a change should be aimed at.

Playbook for finding and fixing an imbalance in a game's in-game economy (inflation, a dominant source, an unused sink, a stalled premium conversion) and shipping the fix as a measurable, reversible experiment. Load when the user asks about currency balance, prices, rewards, or inflation.

Understand and query the Community Pulse feature: sentiment sources, topic trends, player complaints, praise, and representative quotes.

Navigate, read, and search remote configurations — large JSON blobs rendered as a file tree in the dashboard.

Look up deeplink definitions and query interaction, conversion, and LTV analytics for deeplinks.

Understand Gamebeast A/B experiments: the model (groups, weights, unit types, targeting, lifecycle), how to measure results with get_experiment_results and read its statistics, and how to filter or break down insight/funnel queries by experiment group. Load before analyzing any experiment.

Playbook for a two-minute weekly review: what moved in the last 7 days and why, how running experiments are doing, what players are saying, and the one or two things to do next, saved to a dashboard so next week starts from the same view. Load when the user asks for a weekly review, a status summary, or how the project is doing.