feat(cts): conformance test suite and docs for hand and full-body device integrations

.gitattributes

@@ -21,6 +21,10 @@
 # Video assets are tracked by default.
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 
+# Audio assets are tracked by default.
+*.wav filter=lfs diff=lfs merge=lfs -text
+*.mp3 filter=lfs diff=lfs merge=lfs -text
+
 #
 # Exceptions for small source SVGs (webpack-inlined; not binary assets).
 deps/cloudxr/webxr_client/src/icons/*.svg -filter -diff -merge text

README.md

@@ -69,6 +69,10 @@ Our [documentation page](https://nvidia.github.io/IsaacCapture) provides everyth
 - [Quick installation steps](https://nvidia.github.io/IsaacCapture/main/getting_started/quick_start.html)
 - [How to build from source](https://nvidia.github.io/IsaacCapture/main/getting_started/build_from_source.html)
 
+### Check a Device Integration
+
+The CTS records one scripted take from a hand-tracking or full-body device and checks it against the data a retargeter needs; start with the [hand](src/python/isaaccapture/cts/hand/README.md) or [full-body](src/python/isaaccapture/cts/full_body/README.md) guide.
+
 
 ### Install & Run Isaac Lab
 

docs/AGENTS.md

@@ -33,7 +33,9 @@ Images follow `.gitattributes` like everything else, Git LFS included. A checkou
 LFS holds pointer text and Sphinx copies it into the site verbatim, so a page that looks
 right locally can still ship broken artwork. Commit artwork together with the page or data
 that references it. Stage new artwork with a plain `git add` before running pre-commit:
-`git add -N` bypasses the LFS filter, and `check-lfs-pointers` rejects the raw blob.
+`git add -N` bypasses the LFS filter, and `check-lfs-pointers` rejects the raw blob. A clone
+whose hooks pre-commit installed has no LFS pre-push hook, so GitHub declines a push that
+adds artwork; run `git lfs push origin <branch>` first.
 
 ## Device pages stand alone
 
@@ -96,6 +98,14 @@ device page documents it, links belong in the Details panel, and the sentence it
 dropped. The build keeps `company` that narrow by rejecting every field but `name`, `logo`,
 and `logo_dark`.
 
+## Conformance pages name no device maker
+
+`source/conformance/` and its media are vendor-neutral: no glove, suit or tracker company
+or product appears in prose, cards, commands, take names or plugin names, because a named
+example reads as an endorsement. Use a fictional device (Glove X with `glove_x_plugin`,
+Suit X with `body.suit-x`) and say it is a placeholder. Headsets (PICO, Quest) are the
+exception: every take runs on one.
+
 ## Removing cache overrides
 
 Keep cache-removal documentation focused on migration impact. Omit asset-fetch

docs/source/_ext/breakable_literals.py

@@ -0,0 +1,35 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+"""Let long dotted or underscored names in HTML literals wrap after ``.`` and ``_``."""
+
+from __future__ import annotations
+
+import re
+
+from docutils import nodes
+
+BREAKABLE = re.compile(r"[A-Za-z0-9_.]{24,}")
+AFTER_SEPARATOR = re.compile(r"(?<=[._])")
+
+
+# The search index reads the doctree, so the breaks are added only when writing HTML.
+def _visit_literal(self, node: nodes.literal) -> None:
+    type(self).visit_literal(self, node)
+    text = node.astext()
+    if len(node.children) != 1 or not isinstance(node[0], nodes.Text):
+        return
+    if not BREAKABLE.fullmatch(text):
+        return
+    parts = [part for part in AFTER_SEPARATOR.split(text) if part]
+    for index, part in enumerate(parts):
+        if index:
+            self.body.append("<wbr>")
+        self.visit_Text(nodes.Text(part))
+    type(self).depart_literal(self, node)
+    raise nodes.SkipNode
+
+
+def setup(app):
+    app.add_node(nodes.literal, override=True, html=(_visit_literal, None))
+    return {"version": "1.0", "parallel_read_safe": True, "parallel_write_safe": True}

docs/source/conf.py

@@ -67,6 +67,7 @@ def _pip_pin(ref_name, fallback):
     "sphinx_multiversion",
     "sphinx_design",
     "ecosystem_grid",
+    "breakable_literals",
 ]
 
 exclude_patterns = ["build", "_templates", "_data", "_ext", "Thumbs.db", ".DS_Store"]