Find the skills your whole team struggles with in Chorus

**Objective:** Find the recurring weaknesses across the coaching initiative `{initiative-name}`: which scorecard questions score lowest across everyone being reviewed, and what the reviewers actually write about them. Use scores at or below `{score-threshold}` (default 3) as the definition of weak.

**Step 1: Establish the initiative**

Call `chorus-get-scorecards` with `initiative-name.contains` set to `{initiative-name}`. Confirm the exact `initiative-name` and note how many scorecards exist and over what date span. If several distinct initiatives match, list them and stop.

**Step 2: Pull the weak scores**

Call `chorus-get-scorecard-scores` with `initiative-name.contains` set to `{initiative-name}` and `score.max` set to `{score-threshold}`. Collect `question`, `score`, `recipient-name`, `reviewer-name` and `comment`.

**Step 3: Pull the baseline**

Call `chorus-get-scorecard-scores` again with only `initiative-name.contains` set, so you have every score for the initiative. This is what turns a raw count into a rate: a question asked 200 times with 40 weak scores is a different problem from one asked 45 times with 40.

**Step 4: Group the themes**

Group the weak scores by `question`. For each, compute the weak count, the total count, the weak rate and the average score. Then read the `comment` text on the weak scores for that question and pull out the recurring phrasing. Do not invent a theme that only one comment supports.

**Step 5: Report**

Produce:

* **Ranked questions** - a table of question, weak count, total asked, weak rate, average score, sorted by weak rate.
* **What reviewers say** - for the three worst questions, two or three quoted comments each.
* **Who it concentrates on** - whether the weak scores spread across many people or cluster on a few, with the `recipient-name` counts.
* **Reviewer variation** - whether one reviewer scores markedly harder, from the `reviewer-name` split.
* **Recommended intervention** - one thing to change, named, with the evidence behind it.

**Rules & Guidelines:**

* This is read only. Do not call any tool that creates, updates or exports.
* Report a rate alongside every count. A bare count of weak scores is misleading.
* Distinguish a question scoring low from a question being scored harshly by one reviewer. Check the reviewer split before calling it a skills gap.

**Validation:**

* If no scorecards match `{initiative-name}`, list the initiative names that do exist and stop.
* If fewer than 20 scores come back in total, say the sample is too small to rank and report the raw scores instead.