Judge a recurring webinar against real benchmarks

**Objective:** Report how the webinar `{webinar-name}` performed across its sessions in the last `{lookback-days}` days (default 90): who registered, who turned up, how long they stayed and how engaged they were.

**Step 1: Identify the webinar**

Call `goto-webinar-get-webinar` with webinar-name.contains set to `{webinar-name}`. If nothing matches, stop and say so. If more than one distinct webinar key matches, list each match with its webinar name, date and webinar key, and ask the user which one to report on. Do not merge different webinars.

**Step 2: Pull the sessions**

Call `goto-webinar-get-session` with the webinar key and date.from set to today minus the lookback days. For each session capture: date, registrant-count, registrants-attended, percentage-attendance, average-attendance-time, average-attentiveness, average-interest-rating, questions-asked and total-poll-count.

**Step 3: Interpret**

* Attendance: registrants-attended as a share of registrant-count. Roughly 35 to 50 percent is typical for a live webinar; below 30 percent points to a reminder or timing problem.
* Retention: average-attendance-time against the session length (end-time minus start-time). Under half the session means people are leaving early.
* Engagement: average-attentiveness, questions-asked and total-poll-count together. A session with no questions and no polls was a broadcast, not a conversation.
* Trend: whether attendance and attentiveness are rising or falling across the sessions in date order.

**Step 4: Report**

* A one-line verdict.
* A table, one row per session in date order: Date, Registered, Attended, Attendance %, Avg time attended, Attentiveness, Interest rating, Questions, Polls.
* A totals row (sums for counts, averages for rates).
* Two or three recommendations, each tied to a specific number in the table.

**Rules & Guidelines:**

* This prompt is read-only. Never call `goto-webinar-add-registrant` or `goto-webinar-delete-registrant`.
* Check the scale of the percentage fields from the values returned (0 to 1 or 0 to 100) before comparing them to the thresholds above.
* Show a session with zero registrants in the table but leave it out of the averages, and say so.
* Report the numbers as returned. Do not estimate missing values.
* For survey and poll answers in depth, point the user to the summarize-webinar-feedback prompt rather than expanding this report.

**Validation:**

* If the webinar has no sessions in the period, say so and suggest a longer lookback-days.
* If `{webinar-name}` is empty, ask for it before calling any tool.