# Balance, Agility & Conditioning: research notes

Date: 2026-10-25. Scope: portable equipment for training balance and proprioception (rocker boards, wobble/wobble-dome balance trainers, balance pads and discs) and for footwork, speed and reaction drills (agility ladders, low hurdles, cones and markers, reaction balls). One register, one `balance-equipment` slug, taxonomy `supplies > active > balance_equipment`. Lateral-slide conditioning mats ("slide boards") and kids' gymnastics floor beams turned up under the same brand lines and are mentioned in vocabulary and mechanism, but are not given dedicated facets: one brand only, and each is a materially different movement pattern (sideways skating; narrow-beam precision) from the two families the splits are built on. Slacklines are mentioned for the same reason (one primer source, no brand spec sheet reached). Not in scope: free weights, resistance bands, foam rollers, jump ropes, yoga/pilates props and gym benches, which are separate sibling leaves in `catalog.json` under `supplies/active` and already own their own registers (planned or published). If any retailer files a slackline, slide board or gymnastics beam under this leaf, it overlaps lightly with the mechanism text here but is not this register's facet set.

Quote labels: **verbatim** = I fetched the raw page or its JSON myself (curl, then read the text or parsed JSON in Python) and the quoted words are exactly what was in it. **fetched excerpt** = the WebFetch tool read the page and returned a summary, in its own words or lightly paraphrased, sometimes with short quoted fragments. **search excerpt** = a claim that only ever appeared as text was not reachable and the words below are not used for any JSON claim.

WebSearch was exhausted for this session on the first two queries (reported "used its web search budget (200 of 200)"); every source below came from direct WebFetch, curl, a brand's own Shopify `/products.json` endpoint (read and parsed in Python, so counted as verbatim), the NCBI eutils API, and CPSC's SaferProducts.gov REST API. I could not reach: Wirecutter/Verywell-style consumer review roundups (verywellfit.com refused the fetch tool outright), Perform Better and Rogue's category pages (403, and one DNS/connection failure), NIA's and physio-pedia's balance-exercise pages (both now 404 — the NIA page moved when the site was redesigned, physio-pedia blocked the fetch), BOSU's own PDP "Specifications" accordion (present in the markup but empty — the merchant never filled it in; I confirmed this by downloading the raw HTML and reading the panel's content myself rather than trusting a fetch summary), and ASTM's own standard text (astm.org pages 403; I used CPSC's recall database instead to check whether a mandatory standard or recall history exists, and found neither).

## Seed

Seed run output (recorded verbatim):
```
=== Example entry (from an unrelated category) ===
category: kitchen-knives
what this shows: a numeric spec that only makes sense as one side of a trade-off, and a visual grammar borrowed from the hobbyist community that puts the trade-off on one axis.
[steel_hardness facet + trade-off, see schema example file]
=== Persona ===
A textile conservator, who thinks in fibre, construction and how something ages
```

1. **The balance-equipment equivalent of steel hardness is how many directions a board tips in, not a number anyone measures.** Knife steel has one official scale (Rockwell C) that is a genuine trade-off: harder holds an edge longer but chips. Balance gear has no equivalent certified scale — nobody rates a board "62 HRC of wobble" — but Wikipedia's own taxonomy of balance boards is explicit that "what differentiates the five types... is how unstable each of them is, i.e., in how many and in which of the three dimensions of space each board turns and/or sways" (fetched excerpt, quoted). That is the real trade-off: more directions of movement trains the ankle/core harder but is more likely to cause a fall, exactly as "falls from balance boards can break bones, sprain joints, and tear tendons, ligaments and cartilage" (fetched excerpt, quoted, same page). **Survived the data check**: two real products from one brand (ProsourceFit) independently describe themselves at the two ends of this axis — the Wooden Rocker Board says "side or frontal movements" (verbatim) while the Wooden Balance Board says "360-degree rotation allows you to shift your weight in any direction" (verbatim) — so the axis is not just an encyclopedia abstraction, it is how sellers themselves describe different SKUs. Kept as the `instability_style` split and the first trade-off. Dropped: inventing a numeric "instability score" — no source measures one.
2. **The textile conservator would notice the webbing, which is the one soft-goods part of an otherwise hard-goods category.** Agility ladders that are not rigid/hinged are built from flat straps with sliding rungs — ProsourceFit's own copy calls them "flat adjustable rungs" (verbatim) on both its flat and raised ladders — and a slackline is, per Wikipedia, webbing "tensioned significantly less than tightropes," in widths from "2-inch (5 cm)" trick lines down to "5⁄8 or 1 inch (1.6 or 2.5 cm)" long lines (fetched excerpt, quoted). A conservator's instinct — check the stitching at the load-bearing loop, check for UV-fray on the strap, because that is the part that silently loses strength under repeated dynamic load — does not show up as a field on any spec page I reached, and I found no second source discussing ladder-webbing wear specifically. **Partially survived**: kept as the sourced contrast behind the `rung_spacing_style` facet and the second trade-off (fixed/rigid ladders trade customizable spacing for no fraying, no tangling); dropped as its own "webbing condition" facet because nothing backs a buyer-facing claim about how fast it wears.

## Sources and what each supports

### Primers and encyclopedic background

**Wikipedia, "Balance board"**
https://en.wikipedia.org/wiki/Balance_board
- fetched excerpt (quoted): "A rocker board is the most basic and least challenging type of balance board."; "The fulcrum of almost all wobble boards is a semi-sphere or smaller spherical cap..."; "Rather than on a fixed pivot, a rocker-roller's board is placed on a cylindrical roller..."; "A sphere, either an inflatable rubber ball... or a solid polyurethane ball, is the fulcrum..."
- fetched excerpt (quoted): "what differentiates the five types (and their subtypes) is how unstable each of them is, i.e., in how many and in which of the three dimensions of space each board turns and/or sways."
- fetched excerpt (quoted): boards "develop balance, motor coordination skills, weight distribution and core strength" and are used "to prevent sports injuries, especially to the ankle and knee" and "for rehabilitation after injuries."
- fetched excerpt (quoted): "Falls from balance boards can break bones, sprain joints, and tear tendons, ligaments and cartilage."
- Supports: the mechanism, the `instability_style` taxonomy and split, the fall-injury judgement. Confidence high for the taxonomy (it is the one place that states all variants together); medium for exact wording since this came through the fetch tool's summary rather than text I stripped myself.

**Wikipedia, "Slackline"**
https://en.wikipedia.org/wiki/Slackline
- fetched excerpt (quoted): "Slacklines are tensioned significantly less than tightropes or tightwires in order to create a dynamic line which will stretch and bounce like a long and narrow trampoline."; "2-inch (5 cm) slackline kits" for tricklining; "narrow (5⁄8 or 1 inch, 1.6 or 2.5 cm) lines for longline purposes."; "When a person slacklines, they typically use their arms to balance when the center of mass is displaced."
- Supports: the vocabulary note on slacklines and the seed answer on webbing. Confidence medium (fetch-tool summary, not self-read text); not used for any number in the JSON (no `range` set).

### Brand product data (read as raw JSON, not a fetch summary)

For every brand below I pulled the store's own `/products.json` (the Shopify storefront API every one of these sites exposes) or an individual product's `.json` with curl, then parsed it in Python and read `body_html` and the `variants` array myself. Every quote marked verbatim in this section came from that text, not from a WebFetch summary.

**BOSU (bosu.com)**
https://www.bosu.com/products/bosu-elite.json , .../bosu-pro-balance-trainer.json , .../bosu-home-balance-trainer.json , .../sport-balance-trainer-50cm-copy.json , .../k-series.json , .../balance-pad.json
- verbatim, BOSU Elite: "Unlike our Original Balance Trainer, the Elite focuses on resistance training, rather than instability training. With its innovative 'Reactive Dome™' and reinforced platform, the Elite generates powerful elastic resistance..."
- verbatim, Pro/Home/Sport/K-Series Balance Trainer listings: "challenges the entire body with integrated, multi-joint movements"; "designed with six non-slip feet"; "Burst Resistant Smooth Ribbed Dome"; K-Series "8 QUADRANT, GRIP-TEXTURED DOME... BURST-RESISTANT BLADDER... IN THE BOX 1 Cover, 1 Bladder, 1 Platform, PUMP & 2 PLUGS."
- verbatim, Balance Pad: "The flexible TPE foam surface creates gentle instability, forcing your core, feet and ankle muscles to activate with simple movements... great for beginners & older adults to increase body awareness."
- Variant weights (grams, from the `variants` array, verbatim): Pro Balance Trainer 8,391 g; Home Balance Trainer 7,938 g; Sport (50 cm, "a smaller version of the Home Balance Trainer") 5,670 g; K-Series 3,629 g.
- BOSU homepage fetched excerpt (quoted): "both sides utilized" functionality — the platform can be used dome-up or flat-side-up.
- Supports: the balance-trainer family description, the dome/bladder/platform construction, the resistance-vs-instability distinction (kept as a judgement/vocabulary note, not a facet — it is a single-brand marketing distinction and I found no second brand drawing it), the reversible-sides fact (paired with ProsourceFit below). The "Specifications" accordion on every BOSU PDP I checked was empty in the raw HTML (confirmed by downloading and reading it, not just a fetch summary) — BOSU states no dome diameter, inflation PSI or weight capacity anywhere I could reach, which is why this register does not put a BOSU-sourced number in the JSON. Confidence high for what the text says, since I read it myself.

**SKLZ (sklz.com)**
https://www.sklz.com/products/quick-ladder-0124.json , elevation-ladder-0940.json , quick-ladder-pro-1861.json , 6x-hurdles-1860.json , agility-cones-0127.json , reaction-ball-3508.json , agility-trainer-pro.json , mini-cones.json
- verbatim, Quick Ladder: "features 11 heavy-duty plastic rungs that measure 17” wide and 15' long... folds up easily."
- verbatim, Elevation Ladder: "the SKLZ Elevation Ladder is designed to have the flexibility of a flat ladder as well as an elevated ladder... switches between a flat ladder and 4” hurdles within seconds... measures 7' long and each rung measures 26” wide and 15” long. Made of durable yet lightweight PVC plastic."
- verbatim, Quick Ladder Pro: "the ladder is designed without webbing and features rigid, hinged sides... With extensible ends and sides... the low-profile edges limit cleats and shoes from catching."
- verbatim, 6X Hurdles: "designed to stand up to the biggest beatings... bounce-back construction and maintain their shape after being stepped on... comes in a convenient six-pack." (No stated height in the product text itself; the SKLZ collection-page summary called the set "height-adjustable," which is a fetched excerpt, not something I verified in this product's own text, so "adjustable height" for the 6X Hurdles is not used in the JSON.)
- verbatim, Agility Cones: "the 2 inch cones won't break when stepped on... Includes 20 cones – 5 each of yellow, black, white, orange."
- verbatim, Reaction Ball: "the six-sided shape and high-bounce rubber material causes the ball to leap and hop randomly."
- verbatim, Agility Trainer Pro: "unique trapezoid design... set of 10 with a carry handle."
- Supports: the `agility_tool_type` split and values, the `rung_spacing_style` fixed/rigid side of the contrast, the ladder-length facet's short end (7 ft), the hurdle-height facet (4 in, paired with ProsourceFit below).

**ProsourceFit (prosourcefit.com)**
https://www.prosourcefit.com/products/balance-trainer.json , raised-agility-ladder.json , speed-agility-ladder.json , exercise-balance-pad-large.json , exercise-balance-pad.json , wooden-balance-board.json , wooden-rocker-board.json , core-balance-disc.json , slide-board.json , slide-board-pro.json , gymnastics-balance-beam.json
- verbatim, Balance Trainer (dome): "Exercise platform side up or down for challenging balance workouts... Using the convenient air pump, inflate the dome to the desired level for safe use... Made from burst resistant PVC and rubber, the balance ball can hold up to 330 pounds." (330 lb is one brand only; not used as a JSON range because the rule here is two sources for a number, and I found no second brand stating a figure — BOSU's own pages state none at all.)
- verbatim, Raised Agility Ladder: "Easily fold down the 6 rungs to a flat speed ladder... The 4-inch raised rungs of the workout ladder are perfect to increase steps and lift speed... Comes in 3 lengths with flat adjustable rungs: 4m/8 rungs, 6m/12 rungs, 10m/20 rungs."
- verbatim, Speed Agility Ladder: "Comes in 3 lengths with flat adjustable rungs: 4m/8 rungs, 6m/12 rungs, 10m/20 rungs... Personal trainers and therapists will find them useful for rehabilitation as well as improving clients' balance and body control."
- verbatim, Wooden Rocker Board: "Side or frontal movements make a fun challenge... ideal for strengthening and conditioning or physical therapy after surgery."
- verbatim, Wooden Balance Board: "360-degree rotation allows you to shift your weight in any direction and strengthen all of your major muscle groups." Variant weight: 1,860 g ("lightweight three-pound wobble board" in the body text — 1,860 g is about 4.1 lb, so the "three-pound" marketing line and the cart weight disagree by about a pound; neither is used as a JSON number).
- verbatim, Exercise Balance Pad (regular) and (large): "The flexible foam surface forces your core to engage, and your feet and ankle muscles to activate"; "Closed-cell, TPE material"; regular pad "2.4-inch thick," large pad "15x19-inch... 2-inch thick" (thickness not used in the JSON: one brand only, and the large pad is stated as thinner than the regular pad, which reads as an error I cannot resolve against a second source).
- verbatim, Core Balance Disc: "the uneven surface... dual-textured for a non-slip grip... Dimensions: 2.2 lb, 14-in Diameter x 3-in Height (5-in max.)." (One brand; not used as a JSON range.)
- verbatim, Slide Board / Slide Board Pro: "The 6-foot long mat offers smooth skating with bumpers... ideal for cardio and lower body workouts, hockey training, and physical therapy... A textured PVC bottom prevents sliding when the mat is on hard floors."
- verbatim, Gymnastics Balance Beam: "The 9 ft foldable balance beam helps young gymnasts... a wider base... no height drop and a soft landing."
- Supports: the `instability_style` split (rocker vs. wobble, same brand, two SKUs), the balance-pad mechanism (paired with BOSU above), the reversible-sides fact (paired with BOSU above), the `agility_tool_type` split and the adjustable side of `rung_spacing_style`, the ladder-length facet's typical band (4 to 10 m), the hurdle-height facet (4 in, paired with SKLZ above), and the vocabulary notes on slide boards and gymnastics beams (kept out of the facet set; see Scope).

### Re-fetched 2026-10-25, for the blind-check fixes

The blind check (`check.md`) found several claims cited to the wrong single-SKU page on sklz.com or prosourcefit.com. I re-pulled each product's own `.json` with curl and read `body_html` myself (verbatim, same method as the original research) to confirm the real page for each claim, then added a dedicated reference entry per page rather than reusing a generic brand-wide id for a claim it doesn't carry.

**SKLZ Quick Ladder** — https://www.sklz.com/products/quick-ladder-0124 — verbatim: "the Quick Ladder features 11 heavy-duty plastic rungs that measure 17” wide and 15' long." Supports `ladder_length`'s long-model figure (15 ft ≈ 4.6 m). This page never mentions webbing or rigidity, so it is not evidence for `rung_spacing_style`.

**SKLZ Elevation Ladder** — https://www.sklz.com/products/elevation-ladder-0940 — verbatim: "the ladder switches between a flat ladder and 4” hurdles within seconds... The ladder measures 7' long." Supports `hurdle_step_height`'s 4-in figure and `ladder_length`'s short-model figure (7 ft ≈ 2.1 m), and is the sourced basis for the new `agility_tool_type` value "raised ladder."

**SKLZ Reaction Ball** — https://www.sklz.com/products/reaction-ball-3508 — verbatim: "The six-sided shape and high-bounce rubber material causes the ball to leap and hop randomly." Supports the new `agility_tool_type` value "reaction ball or rings."

**SKLZ Agility Trainer Pro** — https://www.sklz.com/products/agility-trainer-pro — verbatim: "features a unique trapezoid design... Coming in a set of 10 with a carry handle." Supports the same new value.

**ProsourceFit Raised Agility Ladder** — https://www.prosourcefit.com/products/raised-agility-ladder — verbatim: "The 4-inch raised rungs of the workout ladder are perfect to increase steps and lift speed... Comes in 3 lengths with flat adjustable rungs: 4m/8 rungs, 6m/12 rungs, 10m/20 rungs." Supports `hurdle_step_height`'s 4-in figure, `ladder_length`'s max (10 m) and the adjustable-webbing half of `rung_spacing_style`. (The rigid/no-webbing half stays correctly cited to SKLZ's Quick Ladder Pro, which does say "designed without webbing and features rigid, hinged sides" — that citation was already right.)

**ProsourceFit Wooden Balance Board** — https://www.prosourcefit.com/products/wooden-balance-board — verbatim: "360-degree rotation allows you to shift your weight in any direction." Supports the wobble side of `instability_style`'s two-sided quote. (The rocker side, "Side or frontal movements," stays correctly cited to the Wooden Rocker Board page — that citation was already right.)

**CPSC SaferProducts.gov** — re-ran all four REST queries myself on 2026-10-25 (`RecallTitle=balance%20board`, `wobble%20board`, `balance%20trainer`, `agility%20ladder`, each `&format=json`): all four returned `[]`. Each now has its own reference entry and url; previously only the "balance board" query had a citation even though the judgement leaned on all four.

I also removed `bosu_site` from `instability_style`'s evidence. Its one cited page, BOSU Elite, says "Unlike our Original Balance Trainer, the Elite focuses on resistance training, rather than instability training" — it does not support "wobbles every way," and actively names itself the non-instability model. No other BOSU page I reached states a 360-degree or multi-direction instability claim either, so the facet now rests on Wikipedia and ProsourceFit's two SKUs only, which do support it directly.

### Regulator

**CPSC, SaferProducts.gov REST recall search**
https://www.saferproducts.gov/RestWebServices/Recall?RecallTitle=balance%20board&format=json (and the same with "wobble board", "balance trainer", "agility ladder")
- verbatim: all four queries returned `[]` — zero recalls on record for any of those terms, checked 2026-10-25.
- Supports: the judgement that no category-specific mandatory safety standard or recall history exists for this equipment, unlike trampolines (ASTM F2276) or playgrounds (ASTM F1487). Confidence medium: an empty recall search is an absence of evidence, not proof a standard does not exist, and I could not reach astm.org (403) to check for a voluntary ASTM standard directly.

### Could not reach

verywellfit.com (fetch tool refused the domain outright), performbetter.com and rogue.com category pages (403 / connection refused), nia.nih.gov's balance-exercises page (404 — URL moved in a site redesign), physio-pedia.com (403/404), astm.org (403), gosports.com and yes4all.com (`/products.json` 404 — wrong guess at their platform, not a block), cdc.gov/steadi (403). NCBI's eutils search for wobble-board/balance-training literature returned mostly unrelated recent sports-science papers under a relevance-free recency sort and no clearly on-topic open-access review turned up in the time available, so no PubMed/PMC source is cited in the JSON.

## Data points used for derived metrics

None. I considered a price-per-piece metric for multi-count sets (cones, hurdles) but dropped it: it is the same arithmetic for any multi-pack good (screws, pencils, cones), which is exactly the generic pattern the brief says to test splits against, and nothing here makes it category-specific the way cost-per-serve or price-per-Wh is in formulations or electronics. `derived_metrics` is left out of the JSON rather than padded.

## Against the existing add-on schema

`zamana-expertise/data/schema-parts/supplies/active/add-ons/balance_equipment.jsonc` currently holds, in full:
```
//Balance, Agility & Conditioning
{
  "sizing": {
    "rope_length": { "value": "number | null", "unit": "enum [m, ft] | null" },
    "rope_diameter": { "value": "number | null", "unit": "enum [mm, in] | null" },
  },
}
```
This is wrong, not just thin. `rope_length`/`rope_diameter` are the jump-rope add-on's fields (`jump_rope.jsonc` has `rope_length` and `added_weight`; this file has `rope_length` and `rope_diameter` instead, which looks like a copy-paste from a rope-ladder or jump-rope template that was never corrected). Nothing this register covers — balance boards, domes, pads, discs, agility ladders, hurdles, cones — has a "rope" with a diameter. None of the balance or agility products in my research mention rope at all except the slackline, which is out of this register's facet set (see Scope). Proposal, nothing edited:

| This register's facet | Proposed kind (AGENTS.md vocabulary) | Why |
|---|---|---|
| `instability_style` | hard filter | Distinct literal values (rocker / wobble-dome / soft pad), always in the product title, buyers filter by it directly. |
| `agility_tool_type` | vertical anchor | The fact this whole leaf is fundamentally understood through — which family of footwork tool a listing is — the way materials are the anchor for furnishings. |
| `rung_spacing_style` | hard filter | Two literal values (adjustable webbing / fixed rigid), stated in brand copy even where not in a discrete spec field; a buyer who wants custom stride length filters on it. |
| `ladder_length` | sizing | Bought-by-number, varies by SKU/variant (4 m / 6 m / 10 m, or 7 ft / 15 ft) — exactly AGENTS.md rule 2's "cable length yes" case. |
| `hurdle_step_height` | sizing | Bought-by-number where it applies (raised-rung and hurdle products state it; flat ladders and boards do not). |

Fields this register considered and is **not** recommending: a `reversible_sides` boolean (BOSU and ProsourceFit both describe flipping the dome between platform-up and dome-up use, but every inflatable dome and most discs work this way, so it filters nothing — it is a one-line marketing fact for `high_signal_markdown`, not a schema field) and a numeric weight-capacity or dome-diameter field (only one brand, ProsourceFit, states either number; BOSU states neither; not dense enough across brands to be a currency field yet, per AGENTS.md rule 4's "if we wouldn't augment it, it isn't currency" test — with one brand supplying it and the category leader supplying nothing, augmenting the rest would mean inventing numbers). `max_user_weight` is already a shared field on `active-standard.jsonc`; nothing here needs a category-specific duplicate of it.

## Branches

- `training_focus` = balance and stability → ask `instability_style`; `agility_tool_type` and `rung_spacing_style` stop mattering. Chain: a product that is a board, dome, pad or disc is never also a ladder, hurdle or cone set (BOSU, ProsourceFit and SKLZ each sell one family or the other, never mixed in a single SKU), so the agility facets have no value to show. SUPPORTED.
- `training_focus` = speed and footwork → ask `agility_tool_type`, then (if ladder) `rung_spacing_style` and `ladder_length`, or (if hurdles) `hurdle_step_height`; `instability_style` stops mattering. Chain: same as above, in reverse; SKLZ's and ProsourceFit's ladder/hurdle/cone lines never state an instability type because there is no unstable surface to describe. SUPPORTED.
- `training_focus` = both → ask `agility_tool_type` and `instability_style` (added to `category.json` 2026-10-25; previously only documented here, never in the JSON — see corrections log). No facet is made irrelevant, since a buyer training both will eventually want both sets of facets across different products.
- `agility_tool_type` = flat ladder → ask `rung_spacing_style` and `ladder_length`; `hurdle_step_height` stops mattering. This leg is now restricted to ladders that never convert to a raised/hurdle mode (no second SKU type folded in). SUPPORTED: SKLZ's Quick Ladder and Quick Ladder Pro state length and webbing/rigid construction and never mention a raised or hurdle mode.
- `agility_tool_type` = raised ladder → ask `rung_spacing_style`, `ladder_length` **and** `hurdle_step_height` (none made irrelevant). This is a new value, split out of the old single "ladder" value on 2026-10-25 (see corrections log) because SKLZ's Elevation Ladder and ProsourceFit's Raised Agility Ladder both fold between a flat ladder and a 4-in raised/hurdle mode, so they genuinely have a step height and were wrongly losing it when classified as plain "ladder." SUPPORTED on both brands' own pages (re-fetched and confirmed verbatim 2026-10-25).
- `agility_tool_type` = hurdles or cones and markers → `rung_spacing_style` and `ladder_length` stop mattering (there are no rungs or ladder length on a standalone hurdle or cone set); `hurdle_step_height` applies only to hurdles. Chain: SKLZ's and ProsourceFit's hurdle and cone products never mention webbing, rungs or ladder length. SUPPORTED.
- `agility_tool_type` = reaction ball or rings → `rung_spacing_style`, `ladder_length` and `hurdle_step_height` all stop mattering. New value added 2026-10-25 (see corrections log) for SKLZ's Reaction Ball and Agility Trainer Pro, which have no rungs, length or step height on their own pages (re-fetched and confirmed verbatim 2026-10-25).

## Split answer mapping

Each split question's offered answers map to the split facet's `values` by the shortened words standing in for the full label: `training_focus` — "balance" -> "balance and stability", "speed and footwork" -> "speed and footwork", "both" -> "both". `instability_style` — the question uses the exact three values verbatim, no shortening. `agility_tool_type` — "flat ladder" -> "flat ladder", "raised ladder" -> "raised ladder", "hurdles" -> "hurdles", "cones" -> "cones and markers", "reaction tool" -> "reaction ball or rings", "mixed set" -> "mixed set".

## Trimmed detail

- BOSU's own "Specifications" accordion is present in the page markup on every PDP I checked but empty — confirmed by downloading the raw HTML myself, not by trusting a fetch summary that might have hallucinated a reason. BOSU is the category's best-known brand and states no dome diameter, inflation pressure or weight capacity anywhere reachable; this is why no BOSU-sourced number appears in the JSON, and why `backfill_cost` for anything size-related is not set to "none."
- ProsourceFit's Wooden Balance Board body copy calls itself a "lightweight three-pound wobble board," but the Shopify variant's own `grams` field says 1,860 g (about 4.1 lb). I do not use either number; this is flagged here as a brand's own internal inconsistency, not a claim I am relying on.
- ProsourceFit's two balance-pad SKUs disagree with their own naming: the "Large" pad is 2.4 in thick and the "Extra Large" pad is only 2 in thick (and larger in footprint, 15x19 in). Thickness is not used in the JSON for this reason, on top of being single-brand.
- ProsourceFit's Balance Trainer (dome) states a 330 lb weight capacity; this is the only number of its kind I found anywhere in this category (BOSU states none). Not used in the JSON; recorded here in case a second brand's spec sheet becomes reachable later.
- SKLZ's collection-page blurb calls the 6X Hurdles "height-adjustable," but the product's own page text never says so. Not used.
- The seed's persona lead (webbing/stitching wear as a hidden soft-goods risk) is plausible and matches how slackline and rope-ladder webbing is built, but no source discusses wear or failure rates for it, so it stays as a one-line judgement/vocabulary note rather than a facet.

## Corrections

### Corrections after the blind check (2026-10-25)

1. Required fix 1 (most severe): `agility_tool_type`'s `values` had no slot for a ladder that converts to a raised, 4-in step mode, so the "ladder" branch wrongly made `hurdle_step_height` universally irrelevant even for the register's own sourced raised/elevation ladders. Split the old "ladder" value into "flat ladder" and "raised ladder"; only "flat ladder" now makes `hurdle_step_height` irrelevant, and "raised ladder" asks it along with `rung_spacing_style` and `ladder_length`. Re-fetched SKLZ's Elevation Ladder and ProsourceFit's Raised Agility Ladder pages to confirm both fold between flat and a 4-in raised mode. Also corrected this file's own false claim, in the old "Branches" section, that the exception was "stated above rather than hidden" — it was not; the JSON had no such value until this fix.
2. Required fix 2: `training_focus`="both" had no branch in `category.json` even though this file documented one. Added the branch (`then_ask`: `agility_tool_type`, `instability_style`; nothing made irrelevant, matching this file's existing description).
3. Required fix 3: the `sklz_site` and `prosourcefit_site` references each pointed to one SKU page but were cited for claims that live on different pages of the same brand. Re-fetched the correct pages and added four new reference entries (`sklz_quick_ladder`, `sklz_elevation_ladder`, `prosourcefit_raised_ladder`, `prosourcefit_wooden_balance_board`), and pointed `ladder_length`, `hurdle_step_height`, `rung_spacing_style` and `instability_style`'s `evidence` arrays at the entry that actually carries each claim. `sklz_site` and `prosourcefit_site` keep only the claims their own cited page (Quick Ladder Pro, Wooden Rocker Board) actually states.
4. Required fix 4: `agility_tool_type`'s `values` had no slot for SKLZ's Reaction Ball or Agility Trainer Pro (trapezoid rings), both in scope per this file's Scope paragraph. Added the value "reaction ball or rings," its own branch (makes `rung_spacing_style`, `ladder_length` and `hurdle_step_height` all irrelevant), and two new reference entries (`sklz_reaction_ball`, `sklz_agility_trainer_pro`), re-fetched and quoted verbatim above.
5. Required fix 5: `ladder_length.range.min` was 2, but the shortest sourced product (SKLZ Elevation Ladder) is 2.1 m; no sourced product is exactly 2 m. Changed min to 2.1.
6. Required fix 6: `instability_style`, `agility_tool_type`, `ladder_length` and `hurdle_step_height` were marked `source: "listing"` on the strength of brand storefronts only, never a third-party marketplace listing. Kept `source: "listing"` (the facts are genuinely in ordinary bullet/title copy, just unverified on a reseller's page) but added a new judgement documenting the gap and appended a one-line caveat to each of the four facets' `backfill_note`.
7. Required fix 7: the `cpsc_saferproducts` reference's single url covered only the "balance board" query, though the judgement leans on all four searches. Re-ran all four queries myself (verbatim, all `[]`); added `cpsc_saferproducts_wobble`, `cpsc_saferproducts_trainer` and `cpsc_saferproducts_ladder` as their own reference entries and added them to the judgement's `evidence`.
8. Not in the Required fixes list, but directly touching the same `instability_style` evidence array being corrected in fix 3: removed `bosu_site` from `instability_style`'s evidence. Part A graded this PARTIAL because the one cited BOSU page (BOSU Elite) explicitly calls itself the non-instability model, and no other BOSU page reached states a 360-degree/multi-direction instability claim. The facet now cites only Wikipedia and ProsourceFit's two SKUs, which do support the two-sided quote directly. `bosu_site` is untouched elsewhere (the dome/platform mechanism, reversible-sides and inflation-tunes-difficulty claims, none of which this blind check flagged).
9. Rejected: Part A item 8 (`training_focus`'s rationale "nobody sells a product that is both a wobble board and a ladder" graded UNCITED) and the single-brand numbers logged in "Trimmed detail" (ProsourceFit's 330 lb dome capacity, pad thickness, disc dimensions) were not in the checker's numbered Required fixes list, so left as-is rather than expanding scope beyond what was asked; `training_focus`'s evidence list is unchanged.