Add kids feed documentation

docs/feed-engine.md

@@ -128,6 +128,16 @@ Hooks can be provided globally when registering the feed or per-execution via
 uses this mechanism so Phase 1 feeds can plug in actor-specific suppression
 logic later without changing the core pipeline.
 
+## Feed configuration overview
+
+The feed registry in `js/app.js` registers each feed with a `defaultConfig` and
+optional `configSchema`. The Kids feed (`kids`) is the first feed to ship with a
+custom schema:
+
+| Feed | Config schema summary |
+| --- | --- |
+| `kids` | `ageGroup` (`toddler`/`preschool`/`early`/`older`), `educationalTags` (string array), `disallowedWarnings` (string array) — see [`docs/feeds/kids-feed.md`](feeds/kids-feed.md) for the full behavior and defaults. |
+
 ## Why-Metadata
 
 `context.addWhy()` records structured audit entries. All built-in stages use it

docs/feeds/kids-feed.md

@@ -0,0 +1,84 @@
+# Kids feed
+
+The Kids feed (`kids`) is a curated feed-engine pipeline that only surfaces
+videos explicitly marked for kids and then re-ranks them using age-appropriate
+scoring. It lives in the feed registry in `js/app.js` and runs through the
+feed engine stages documented below.
+
+## Hard include/exclude rules
+
+The Kids feed hard-gates content before scoring:
+
+- **Must be marked for kids**: `video.isForKids` must be `true` or the item is
+  dropped.
+- **Never allow NSFW**: `video.isNsfw === true` is always excluded.
+- **Drop invalid or blocked items**: Videos marked `invalid`, or rejected by the
+  standard `shouldIncludeVideo` checks (privacy, access control, blacklists,
+  author blocks) are excluded.
+- **Content warnings**: Any `video.contentWarning` entry that matches the
+  `disallowedWarnings` list removes the item from the feed.
+
+The default `disallowedWarnings` list is:
+`nudity`, `sexual`, `graphic-violence`, `self-harm`, and `drugs`. Feed config or
+runtime overrides can replace or extend this list.
+
+## Age-group definitions + duration bounds
+
+Age groups tune the scoring defaults (preferred tags, educational tags, and
+maximum duration). The feed defaults to the **preschool** profile unless
+`config.ageGroup` or a runtime override is supplied.
+
+| Age group | Max duration | Preferred tags (default) | Educational tags (default) |
+| --- | --- | --- | --- |
+| `toddler` | 5 minutes | toddler, baby, nursery, colors, shapes, lullaby | abc, numbers, counting, learning, alphabet |
+| `preschool` | 10 minutes | preschool, kindergarten, storytime, letters, phonics | counting, alphabet, reading, learning, math |
+| `early` | 15 minutes | early, kids, reading, science, animals, art | science, math, reading, history, geography |
+| `older` | 20 minutes | tween, teens, tutorial, stem, coding, music | stem, coding, history, geography, tutorial |
+
+Duration affects the **age-appropriateness** score. Videos longer than the
+age-group maximum are down-ranked proportionally.
+
+## Moderation thresholds + multi-category report policy
+
+The Kids feed applies stricter moderation thresholds than the general feeds:
+
+- **Blur**: trusted report count ≥ **1**.
+- **Hide**: trusted report count ≥ **1**.
+- **Hide (trusted mutes)**: trusted mute count ≥ **1**.
+
+Moderation runs **three separate stages** for report categories: `nudity`,
+`violence`, and `self-harm`. Each stage pulls the trusted report count for its
+own category, so **counts do not accumulate across categories**. This means a
+report spike in one category only affects that category’s moderation stage.
+Use the why-trail (below) to inspect which category triggered an action.
+
+## Explainability (`addWhy`)
+
+The Kids feed relies on `context.addWhy()` so future UI surfaces can explain
+filtering and scoring decisions:
+
+- **Audience filter** reasons: `not-for-kids`, `nsfw`, `invalid`, `blacklist`,
+  and `content-warning` (includes the matched warning).
+- **Moderation** reasons: `blur` / `autoplay-block` with the report category,
+  plus `trusted-mute` / `viewer-mute` when applicable.
+- **Scoring** reasons: the dominant positive scoring component (`age-appropriateness`,
+  `educational-boost`, `author-trust`, `popularity`, or `freshness`) is recorded
+  alongside the computed score.
+
+These entries are returned in `metadata.why` whenever the feed engine executes.
+
+## Feed configuration schema
+
+The Kids feed exposes a config schema via the feed engine registry:
+
+```json
+{
+  "ageGroup": "preschool",
+  "educationalTags": [],
+  "disallowedWarnings": ["nudity", "sexual", "graphic-violence", "self-harm", "drugs"]
+}
+```
+
+Use `ageGroup` to select the default scoring profile, `educationalTags` to
+override the default educational tag list for that age group, and
+`disallowedWarnings` to adjust the hard content-warning exclusions.