diff --git a/src/skillspector/nodes/analyzers/static_patterns_tool_misuse.py b/src/skillspector/nodes/analyzers/static_patterns_tool_misuse.py
index eb7fc60d5..e5240a58e 100644
--- a/src/skillspector/nodes/analyzers/static_patterns_tool_misuse.py
+++ b/src/skillspector/nodes/analyzers/static_patterns_tool_misuse.py
@@ -66,6 +66,9 @@
 _RUNTIME_SHELL_PARAMETER_SENTINEL = "\ue002"
 _SIMPLE_BRACED_PARAMETER_RE = re.compile(r"\$\{(?:[A-Za-z_][A-Za-z0-9_]*|[0-9]+|[@*#?$!-])\}")
 _SHELL_DELIMITER_WORD_RE = re.compile(r"[^$'\"`\\(){}<>#;|&!\s]++")
+# A ``#`` that starts a word begins a shell comment. Quoting is not tracked, so
+# ``echo "a # b"`` also matches; callers only use this to restrict ownership.
+_SHELL_COMMENT_START_RE = re.compile(r"(?<![^\s;&|()<>])#")
 _PERL_LITERAL_PRINT_RE = re.compile(
     r"^[ \t]*+print\b(?:[ \t]++(?:STDOUT|STDERR)\b)?[ \t]*+(?P<paren>\()?[ \t]*+"
     r"(?P<literal>\"(?:\\[\\\"'nrt]|[^\\\"$@`\r\n])*+\""
@@ -1421,9 +1424,14 @@ def _parse_shell_command_word(
     backtick_end_cache: dict[int, int | None] | None = None,
     *,
     check_runtime: Callable[[], None] | None = None,
+    owned_word_positions: set[int] | None = None,
+    enclosing_delimiter: str | None = None,
 ) -> _ShellCommandWord | None:
     runtime_check = check_runtime or (lambda: None)
     output: list[str] = []
+    wrapper_quote = enclosing_delimiter or (
+        _command_wrapper_quote(content, start) if content[start] in "$`" else None
+    )
     quote: str | None = None
     ansi_c_quote = False
     dynamic = False
@@ -1437,6 +1445,8 @@ def _parse_shell_command_word(
         character = content[cursor]
         if quote is not None:
             if character == quote:
+                if owned_word_positions is not None:
+                    owned_word_positions.add(cursor)
                 quote = None
                 ansi_c_quote = False
             elif character == "\\" and ansi_c_quote:
@@ -1598,6 +1608,11 @@ def _parse_shell_command_word(
                 cursor = parameter_end
                 continue
         elif character == "`":
+            if enclosing_delimiter == "`":
+                # The enclosing substitution ends here; this is not an opener.
+                if owned_word_positions is not None:
+                    owned_word_positions.add(cursor)
+                break
             substitution_end = _skip_backtick_substitution(
                 content,
                 cursor,
@@ -1630,6 +1645,8 @@ def _parse_shell_command_word(
             cursor = substitution_end
             continue
         elif character in "'\"":
+            if character == wrapper_quote:
+                break
             quote = character
         elif character == "\\" and cursor + 1 < limit:
             if content[cursor + 1] == "\n":
@@ -1726,6 +1743,9 @@ def _has_shell_command_word_exhaustion(
 ) -> bool:
     """Find candidate command words whose deterministic parse hit a safety bound."""
     parsed_through = 0
+    # Completed words own closing quotes, never executable expansion starts.
+    # Inner commands remain independent candidates; never suppress their bodies.
+    owned_word_positions: set[int] = set()
     parameter_end_cache: dict[int, _ParameterExpansionEnd] = {}
     substitution_end_cache: dict[int, int | None] = {}
     backtick_end_cache: dict[int, int | None] = {}
@@ -1734,17 +1754,23 @@ def _has_shell_command_word_exhaustion(
     for candidate in _SHELL_COMMAND_WORD_START_RE.finditer(content):
         check_runtime()
         start = candidate.start()
+        if start in owned_word_positions:
+            continue
         if structural_quote_closers is not None and start in structural_quote_closers:
             continue
         json_string_start = structural_quote_openers is not None and (
             start in structural_quote_openers or start - 1 in structural_quote_openers
         )
+        assignment_quote = start > 0 and content[start - 1] == "=" and content[start] in "'\""
         if start < parsed_through or (
-            not json_string_start and not _is_shell_command_word_start(content, start)
+            not json_string_start
+            and not assignment_quote
+            and not _is_shell_command_word_start(content, start)
         ):
             continue
         if _has_quoted_assignment_prefix(content, start):
             continue
+        candidate_word_positions: set[int] = set()
         parsed = _parse_shell_command_word(
             content,
             start,
@@ -1752,6 +1778,10 @@ def _has_shell_command_word_exhaustion(
             substitution_end_cache,
             backtick_end_cache,
             check_runtime=check_runtime,
+            owned_word_positions=candidate_word_positions,
+            enclosing_delimiter=(
+                _assignment_value_wrapper(content, start) if assignment_quote else None
+            ),
         )
         if parsed is None:
             substitution_start = (
@@ -1789,6 +1819,21 @@ def _has_shell_command_word_exhaustion(
             if unresolved_end - start > _SHELL_COMMAND_WORD_CHARS:
                 return True
             continue
+        # Comments end at a newline, and a quote after ``=`` in prose or host
+        # source may be mis-paired. Such a word cannot own a later line.
+        confined_word = "\n" in content[start : parsed.end] and (
+            assignment_quote
+            or _SHELL_COMMENT_START_RE.search(content, content.rfind("\n", 0, start) + 1, start)
+            is not None
+        )
+        if not confined_word:
+            # A quote that opens a runtime-selected word stays an independent
+            # candidate, so a mis-paired claim cannot hide its operands.
+            owned_word_positions.update(
+                position
+                for position in candidate_word_positions
+                if content[position + 1 : position + 2] not in ("$", "`")
+            )
         # Only executable nested substitutions retain independent command
         # positions. A plain dynamic data argument still owns its inner bytes;
         # revisiting those as commands would turn quoted printf data into code.
@@ -1805,8 +1850,12 @@ def _has_shell_command_word_exhaustion(
             or "$" not in raw_word
             or not any(marker in raw_word for marker in ("$(", "`"))
             or simple_backtick_parameter
-        ):
+        ) and not (assignment_quote and confined_word):
             parsed_through = max(parsed_through, parsed.end)
+        if assignment_quote:
+            # An assignment value never names the command, and main never
+            # parsed this position, so it only claims ownership.
+            continue
         if parsed.limited:
             return True
         if parsed.dynamic and may_have_destructive_outer_operands:
@@ -2089,6 +2138,19 @@ def _command_wrapper_quote(content: str, command_start: int) -> str | None:
     return content[command_start - 1] if backslashes % 2 == 0 else None
 
 
+def _assignment_value_wrapper(content: str, value_start: int) -> str | None:
+    """Return an unescaped quote or backtick directly enclosing an assignment."""
+    name_start = value_start - 1
+    if name_start > 0 and content[name_start - 1] == "+":
+        name_start -= 1
+    while name_start > 0 and (
+        content[name_start - 1] == "_"
+        or (content[name_start - 1].isascii() and content[name_start - 1].isalnum())
+    ):
+        name_start -= 1
+    return _command_wrapper_quote(content, name_start)
+
+
 def _perl_literal_print_shell_text(
     content: str,
     check_runtime: Callable[[], None],
diff --git a/tests/nodes/analyzers/test_shell_assignment_quotes.py b/tests/nodes/analyzers/test_shell_assignment_quotes.py
new file mode 100644
index 000000000..ef664f9b8
--- /dev/null
+++ b/tests/nodes/analyzers/test_shell_assignment_quotes.py
@@ -0,0 +1,150 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+
+import json
+from pathlib import Path
+
+import pytest
+from typer.testing import CliRunner
+
+from skillspector.cli import app
+from skillspector.nodes.analyzers.static_patterns_tool_misuse import has_bounded_parse_exhaustion
+
+_COMMENT_PADDING = "\n" + "#" * 6000 + "\n"
+_APOSTROPHE_PADDING = "\n" + "y = 'b'\n" * 800
+
+
+@pytest.mark.parametrize(
+    "source",
+    [
+        'out="$(date)"',
+        'v="a (b)"',
+        'v="|$a|"',
+        'printf "%s" "$(f "$x")"',
+        '[ "$(f)" = true ]',
+        'cat <<< "$(f)"',
+        'v="a \\" (b)"',
+        'v="a \\" |$a|"',
+    ],
+)
+def test_complete_shell_quotes_do_not_consume_following_padding(source: str) -> None:
+    assert not has_bounded_parse_exhaustion(source + "\n" + "#" * 6000 + "\n", lambda: None)
+
+
+@pytest.mark.parametrize(
+    "source", ['x="$(unterminated', 'x="unclosed', '"$(printf "%s" "$x")" -rf /']
+)
+def test_unresolved_shell_words_still_exhaust(source: str) -> None:
+    assert has_bounded_parse_exhaustion(source + "\n" + "#" * 6000 + "\n", lambda: None)
+
+
+def test_assignment_ownership_retains_nested_command_budget() -> None:
+    nested_word = "r" * 4097
+    source = f'out="$({nested_word} -rf /)"'
+    assert has_bounded_parse_exhaustion(source, lambda: None)
+
+
+def test_failed_word_parse_does_not_claim_following_command() -> None:
+    source = 'Test-Path "$($_.FullName)\\cli-path"; $($CMD) -rf /'
+    assert has_bounded_parse_exhaustion(source, lambda: None)
+
+
+@pytest.mark.parametrize(
+    "source", ['Run "$(resolve_tool) -rf /"', '# see "notes\n$(resolve_tool) -rf /\n# "']
+)
+def test_quoted_runtime_commands_keep_fail_closed_coverage(source: str) -> None:
+    from skillspector.inspection_ledger import LedgerOutcome, LedgerReason
+    from skillspector.nodes.analyzers import static_patterns_tool_misuse, static_runner
+
+    assert has_bounded_parse_exhaustion(source, lambda: None)
+    path = "script.sh" if source.startswith("#") else "SKILL.md"
+    result = static_runner.run_static_patterns_with_ledger(
+        {"components": [path], "local_file_cache": {path: source}, "file_cache": {path: source}},
+        [static_patterns_tool_misuse],
+    )
+    assert any(
+        e["outcome"] == LedgerOutcome.PARTIAL
+        and e["reason_code"] == LedgerReason.STATIC_PARSE_LIMIT
+        for e in result["inspection_ledger"]
+    )
+
+
+@pytest.mark.parametrize("file_type", ["shell", "markdown", "python"])
+def test_non_shell_unclosed_assignment_remains_conservative(file_type: str) -> None:
+    source = 'name="value' + "\n" + "text " * 1200
+    assert has_bounded_parse_exhaustion(source, lambda: None, file_type=file_type)
+
+
+@pytest.mark.parametrize(
+    ("source", "file_type"),
+    [
+        ('# use "$(cmd)" here', "shell"),
+        ('```bash\nout="$(date)"\n```', "markdown"),
+        ('x="$(printf \'%s\' "$X")"', "shell"),
+        ("model_count=\"$(printf '%s\\n' \"$j\" | jq -r 'length')\"", "shell"),
+    ],
+    ids=["issue-comment-form", "markdown-fence", "printf-value", "printf-pipeline-value"],
+)
+def test_benign_quoted_values_do_not_consume_comment_padding(source: str, file_type: str) -> None:
+    assert not has_bounded_parse_exhaustion(
+        source + _COMMENT_PADDING, lambda: None, file_type=file_type
+    )
+
+
+@pytest.mark.parametrize(
+    ("source", "file_type"),
+    [
+        ('# Keep `X="${X}"` so X\'s value wins.', "python"),
+        ("# Keep `X='${X}'`, so X's value wins.", "python"),
+        ('Use `X="${X}"`. That\'s the self-reference.', "shell"),
+    ],
+    ids=["double-quoted", "single-quoted", "prose"],
+)
+def test_backtick_enclosed_assignment_stops_at_its_closing_backtick(
+    source: str, file_type: str
+) -> None:
+    assert not has_bounded_parse_exhaustion(
+        source + _APOSTROPHE_PADDING, lambda: None, file_type=file_type
+    )
+
+
+@pytest.mark.parametrize(
+    ("source", "file_type"),
+    [
+        ('ls # see "notes\n"$(resolve_tool)" -rf /\n# "', "shell"),
+        ('Set name="value in prose.\n\n```bash\n"$(resolve_tool)" -rf /\n```\n"\n', "markdown"),
+        ('x = f(name="value)\n"$(resolve_tool)" -rf /\n"\n', "python"),
+        ('Set name="value\n$T -rf /\n"\n', "markdown"),
+        ('name="a "$(resolve_tool)" -rf /"', "shell"),
+    ],
+    ids=[
+        "trailing-comment",
+        "prose-assignment-before-fence",
+        "host-language-assignment",
+        "prose-assignment-parameter",
+        "same-line-runtime-opener",
+    ],
+)
+def test_misaligned_quote_claims_keep_runtime_commands_partial(source: str, file_type: str) -> None:
+    assert has_bounded_parse_exhaustion(source, lambda: None, file_type=file_type)
+
+
+def test_issue_628_bundle_reports_complete_coverage(tmp_path: Path) -> None:
+    (tmp_path / "SKILL.md").write_text(
+        "---\nname: repro\ndescription: Minimal repro.\n---\nRun `scripts/run.sh`.\n",
+        encoding="utf-8",
+    )
+    padding = "".join(f"# padding line {index} " + "." * 50 + "\n" for index in range(90))
+    (tmp_path / "scripts").mkdir()
+    (tmp_path / "scripts" / "run.sh").write_text(
+        '#!/bin/bash\nout="$(date)"\necho "$out"\n' + padding, encoding="utf-8"
+    )
+
+    result = CliRunner().invoke(app, ["scan", str(tmp_path), "--format", "json", "--no-llm"])
+
+    # Exit 1 is the documented high-risk verdict, not a scan execution failure.
+    assert result.exit_code in {0, 1}, result.output
+    report = json.loads(result.output)
+    assert report["analysis_completeness"]["status"] == "complete"
+    assert report["analysis_completeness"]["ledger_exceptions"] == []
+    assert not any(issue["id"] == "AE1" for issue in report["issues"])