docs-question-clustererlisted

# docs-question-clusterer — Group every user question and label what's missing `docs-gap-finder` already exists and finds **what to write next** from the demand signals that came back empty — searches that returned nothing, questions the assistant could not answer — set against what readers ask for most. This skill is **deeper**: - It also includes the questions the assistant **did** answer, not only the ones it failed on. - For each cluster, it computes a `coverage_score` against the doc graph — *did an existing page directly answer this?* If yes, the cluster becomes `ai_chat_failure` (doc exists, chat can't surface it = retrieval/prompt problem), not `content_gap` (write a new doc). This distinction matters because the two failures need different actions: | Distinction | Action | |---|---| | `content_gap` | `invoke_skill: docs-create` with a draft outline. | | `ai_chat_failure` | `invoke_skill: docs-tune-ai-chat` — fix retrieval/prompt, not content. | ## When to run - Bi-weekly. - After any AI chat configuration change. - After writing new docs — verify the new docs reduced AI chat failures. ## Workflow Standard four-stage docs-insights pipeline. Slice = `questions`. See [`docs-utm-analyzer`](../docs-utm-analyzer/SKILL.md) for canonical step-by-step. - **Collector:** every question readers put to the assistant — the ones it answered and the ones it could not — plus the answers readers rated badly, the searches that returned nothing, and the queries readers run most