Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@scripts/build_site.py`:
- Around line 163-174: The file reads for SKILL.md, AGENTS.md, and FAQ.md use
Path.read_text() without an explicit encoding; update each call (the expressions
ROOT / "SKILL.md").read_text(), (ROOT / "AGENTS.md").read_text(), and (ROOT /
"FAQ.md").read_text() to pass encoding="utf-8" so files are read consistently
across platforms (preserve the existing SKILL.md YAML-stripping logic using
re.sub on skill).
- Around line 350-352: The loop that writes files (iterating over files, using
DOCS and calling path.write_text(content)) doesn't ensure the target directory
exists, causing FileNotFoundError; before writing each file (or once before the
loop) ensure the directory exists by creating the parent directory (use
DOCS.mkdir(parents=True, exist_ok=True) or path.parent.mkdir(parents=True,
exist_ok=True)) so path.write_text(content) can succeed without errors.

In `@scripts/check_site_assets.py`:
- Around line 24-32: Unused import build_sitemap_xml is present in the import
tuple; remove build_sitemap_xml from the from-import list (the import that
currently brings in DOCS, build_index_md, build_llms_full_txt, build_llms_txt,
build_robots_txt, build_sitemap_xml, _version) so the module no longer imports
an unused symbol and the script remains consistent with skipping sitemap.xml
content comparison.
- Around line 48-51: The checker uses path.read_text() which defaults to locale
encoding and can cause false mismatches; update the comparison to call
path.read_text(encoding="utf-8") and likewise update any write_text() and
read_text() usages in build_site.py to use encoding="utf-8" so generated files
are written and read deterministically; ensure the logic around stale.append(f" 
{name}: missing...") and the content comparison still uses the same UTF-8
decoded string for equality checks.

In `@TESTING.md`:
- Line 370: Update the "Validation Scripts" section to match the lint table:
include check_site_assets.py in the earlier scripts inventory (the section that
lists the "Eleven scripts") and adjust the count/description accordingly so the
list and the lint table (which already shows check_site_assets.py) are
consistent; ensure the filenames check_conformance.py, check_examples.py,
check_version_sync.py, check_spec_examples.py, check_readme_examples.py,
check_skill_examples.py, check_faq_examples.py, check_html_examples.py,
check_site_assets.py, and check_licenses.py appear in the inventory in the same
order/format as the lint table.

---

Outside diff comments:
In `@CONTRIBUTING.md`:
- Around line 74-88: The CONTRIBUTING.md pre-commit hooks list omits the
site-asset documentation validation workflow; update the section that lists "21
hooks" to explicitly mention the site asset/doc validation step and the scripts
involved (build_site.py and scripts/check_site_assets.py or
check_site_assets.py), and add a short guidance sentence that contributors must
run pre-commit or the site asset validation scripts (and not skip
doc-validation) after README/docs edits; reference the exact script names
(build_site.py, check_site_assets.py) and the README/docs change expectation so
readers know which tools to invoke.