mirror of
https://github.com/jhd3197/ServerKit.git
synced 2026-11-02 10:07:29 +09:00
Juan DenisandClaude Opus 4.8
28cece380f
Plan 55 Phase 1. The sweep found zero vulnerabilities: every raw-HTML sink is safe by construction (escape-first SQL/file highlighters, sanitizeSvgInner for extension icons, the escape-then-allowlist markdown renderer) and the backend has no |safe/Markup(/render_template_string. To keep it that way: scripts/check-html-sinks.mjs (wired into npm run lint) and a Python mirror test (test_html_sink_sweep.py, runs in the backend suite) fail on any raw-HTML sink that neither references an allowlisted sanitizer nor carries a 'sink-safe:' annotation. Standardized annotations on the three variable-fed sinks (markdown, SQL editor, marketplace icon) and documented the rule in SECURITY.md. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
62 lines
2.5 KiB
Python
62 lines
2.5 KiB
Python
"""HTML-sink sweep guard — proving test (plan 55 Phase 1, D4).
|
|
|
|
Mirrors scripts/check-html-sinks.mjs in Python so the sweep stays swept inside
|
|
the backend suite (the authoritative gate) even where node isn't available: every
|
|
raw-HTML sink in frontend/src or backend/app must reference an allowlisted
|
|
sanitizer/escaper or carry a `sink-safe:` annotation. A new unannotated sink
|
|
fails this test.
|
|
"""
|
|
import os
|
|
import re
|
|
|
|
_REPO_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
|
|
|
|
FRONTEND_SINK = re.compile(
|
|
r'dangerouslySetInnerHTML|\.innerHTML\s*=[^=]|insertAdjacentHTML'
|
|
r'|new\s+Function\s*\(|\beval\s*\(')
|
|
BACKEND_SINK = re.compile(r'\|\s*safe\b|\bMarkup\s*\(|render_template_string')
|
|
ALLOWLISTED_SANITIZERS = re.compile(r'sanitizeSvgInner|renderMarkdownToHtml|highlightLine|hlSql')
|
|
ANNOTATION = re.compile(r'sink-safe:')
|
|
_COMMENT_PREFIXES = ('//', '*', '/*', '#')
|
|
|
|
|
|
def _walk(root, exts):
|
|
for dirpath, dirnames, filenames in os.walk(root):
|
|
dirnames[:] = [d for d in dirnames
|
|
if d not in ('node_modules', 'dist', '__pycache__')]
|
|
for name in filenames:
|
|
if name.endswith(exts):
|
|
yield os.path.join(dirpath, name)
|
|
|
|
|
|
def _scan(path, sink_re, allow_sanitizer):
|
|
with open(path, 'r', encoding='utf-8', errors='replace') as f:
|
|
lines = f.read().split('\n')
|
|
violations = []
|
|
for i, line in enumerate(lines):
|
|
if not sink_re.search(line):
|
|
continue
|
|
if line.strip().startswith(_COMMENT_PREFIXES):
|
|
continue # a comment that merely mentions a sink keyword
|
|
window = '\n'.join(lines[max(0, i - 3):i + 1])
|
|
cleared = (allow_sanitizer and ALLOWLISTED_SANITIZERS.search(line)) \
|
|
or ANNOTATION.search(window)
|
|
if not cleared:
|
|
violations.append(f'{os.path.relpath(path, _REPO_ROOT)}:{i + 1} {line.strip()}')
|
|
return violations
|
|
|
|
|
|
def test_no_unannotated_html_sinks():
|
|
violations = []
|
|
fe = os.path.join(_REPO_ROOT, 'frontend', 'src')
|
|
for path in _walk(fe, ('.js', '.jsx', '.ts', '.tsx')):
|
|
violations += _scan(path, FRONTEND_SINK, allow_sanitizer=True)
|
|
be = os.path.join(_REPO_ROOT, 'backend', 'app')
|
|
for path in _walk(be, ('.py', '.html')):
|
|
violations += _scan(path, BACKEND_SINK, allow_sanitizer=False)
|
|
|
|
assert not violations, (
|
|
'Unannotated raw-HTML sink(s) — each must reference a sanitizer/escaper '
|
|
'or carry a `sink-safe:` comment (see scripts/check-html-sinks.mjs):\n'
|
|
+ '\n'.join(violations))
|