docs-engagement-analyzerlisted

# docs-engagement-analyzer — Tell interest from confusion A user spending 5 minutes on a page can mean two opposite things: 1. **Deep interest** — pricing, billing, AI chat setup. They're reading carefully because the decision is important. _Good._ 2. **Stuck** — they cannot figure out what the page is telling them. _Bad._ The only reliable disambiguator is **negative feedback** on the same page. This skill combines dwell time and feedback to label every long-dwell page as `engagement_signal` (info) or `engagement_problem` (high severity). ## When to run - Monthly — engagement patterns are slow-moving. - After major content rewrites — verify the rewrite improved things. - When negative feedback on pages spikes. ## Workflow Standard four-stage docs-insights pipeline. Slice = `engagement`. See [`docs-utm-analyzer`](../docs-utm-analyzer/SKILL.md) for the canonical step-by-step. Differences for this skill: 1. **Collector slice:** `engagement` — per-page traffic totals, the raw reading events needed to compute dwell p50/p90 per page, and the feedback readers left on each page. Any other per-page attention signal available (scroll depth, repeat reads) belongs in the same slice. 2. **Clusterer:** groups by page path, computes dwell z-score against site median, joins with feedback counts. 3. **Reporter input:** `SKILL: docs-engagement-analyzer`, `SKILL_VERSION: 1.0.0`. ## Decision matrix (clusterer applies; reporter surfaces) | Dwell time | Negative feedback count | Finding