Which user behaviors are invisible today — and the tracking plan that makes the five questions that matter answerable.
Free & open · no signup · read-only — it ends by asking · nothing leaves your machine
You are working inside this repo. Mission: determine which user behaviors are invisible today, and produce the tracking plan that makes the five questions that matter answerable.
Read-only pass. Your only write is the report file.
Phase 2 sweeps the codebase through every one of these, citing file and line for each finding.
key actions (core loop, aha moment, sharing) emitting no event
stages of signup → activation with no instrumentation between them
inconsistent event names (tense, casing, verb-noun order) making analysis painful
events missing the properties needed to segment (plan, source, count, state)
errors and abandonments untracked; only successes recorded
events that fire but nothing consumes, or duplicates measuring the same act differently
personal data flowing into analytics payloads that shouldn't
One structured report at the repo root — or in reports/, if you keep one — the same shape every time, ready for a teammate — or the optional Studio — to act on.
Every Goal Prompt follows the same four steps, so results are consistent and repeatable — no matter which one you run.
Infer the 5 questions this team most needs answered (from the product's stage and model): e.g.
Invisible behaviors — key actions (core loop, aha moment, sharing) emitting no event.
Judge every gap against the 5 questions.
Create ANALYTICS.md at repo root.
Copy it and paste it into your agent inside the repo you want checked.
Install the goal plugin once — two commands — then just type /goal:analytics-coverage.
/plugin marketplace add GhostlyGawd/goal-prompts/plugin install goal@goal-promptsOr install only this Goal Prompt as /goal-analytics-coverage:
curl -fsSL https://goal-prompts.vercel.app/install | BRIEF=20 shLet an agent fetch it mid-conversation, or pull the raw Goal Prompt by URL.
https://goal-prompts.vercel.app/raw/20.mdNothing hidden — this is the whole Goal Prompt, verbatim. Read it in a minute, edit it, or copy it as-is.
# Goal: Analytics Coverage Audit You are working inside this repo. Mission: determine which user behaviors are invisible today, and produce the tracking plan that makes the five questions that matter answerable. Read-only pass. Your only write is the report file. ## Phase 1 — Establish questions and current state - Infer the 5 questions this team most needs answered (from the product's stage and model): e.g. where does activation fail, which features retain, what precedes churn/upgrade. - Inventory current tracking: which events exist, where they fire, what tool receives them. - Trace one key user flow and list every meaningful action along it. ## Phase 2 — Audit through 7 lenses 1. **Invisible behaviors** — key actions (core loop, aha moment, sharing) emitting no event 2. **Funnel blindness** — stages of signup → activation with no instrumentation between them 3. **Naming chaos** — inconsistent event names (tense, casing, verb-noun order) making analysis painful 4. **Context poverty** — events missing the properties needed to segment (plan, source, count, state) 5. **Failure blindness** — errors and abandonments untracked; only successes recorded 6. **Zombie events** — events that fire but nothing consumes, or duplicates measuring the same act differently 7. **Privacy leaks** — personal data flowing into analytics payloads that shouldn't ## Phase 3 — Curate - Judge every gap against the 5 questions: track what answers them, cut what doesn't - Fewer well-propertied events beat many shallow ones ## Phase 4 — Report Create `ANALYTICS.md` at repo root: 1. **The 5 questions** — and an honest can/can't-answer-today verdict for each 2. **Tracking plan** — table: event name · trigger (file/component) · properties · question it serves 3. **Naming convention** — the rule, plus renames for existing events 4. **Privacy fixes** — payloads to scrub 5. **Implementation order** — sequenced by question priority Start the report with today's date. If `ANALYTICS.md` already exists from a previous run, read it first and lead with what changed since. ## Rules - Every event must serve a named question — no tracking for tracking's sake - Properties are where the insight lives; name them explicitly - No product events to instrument in this repo? Say so in a one-paragraph null report and stop — a null result is a valid finding. - If a `reports/` directory exists at the repo root, write the report there instead of the root. - Before asking, present the top findings as a ranked list in plain words - Report only — end by asking which events to implement
reports/ directory exists at the repo root, write the report there instead of the root.Missing indexes, absent constraints, JSON blobs hiding structure — with a migration plan sequenced safest-first.
Trace where personal data enters, spreads, and lingers — then run the delete-request test: could you honor one today?
Mine the git history for churn, bug magnets, coupled files, and bus-factor silos — predict where the next bug lands.
Schema and data migrations under real traffic — table locks, reversibility, timed-out backfills, and the deploy ordering that turns a change into an outage.