claude-dev-env 1.72.0 → 1.73.0

package/hooks/blocking/code_rules_docstrings.py

@@ -1,6 +1,7 @@
 """Google-style docstring presence and docstring Args-versus-signature checks."""
 
 import ast
+import re
 import sys
 from pathlib import Path
 
@@ -16,29 +17,48 @@ from code_rules_shared import (  # noqa: E402
     _walk_skipping_nested_functions,
     _walk_skipping_type_checking_blocks,
     is_hook_infrastructure,
+    is_strict_test_file,
     is_test_file,
 )
 
 from hooks_constants.blocking_check_limits import (  # noqa: E402
     ALL_DOCSTRING_EXCLUSIVE_SCOPE_PHRASES,
     ALL_DOCSTRING_EXEMPT_DECORATOR_NAMES,
+    ALL_DOCSTRING_FILE_REFERENCE_SUFFIXES,
     ALL_DOCSTRING_IMPLICIT_INSTANCE_PARAMETER_NAMES,
     ALL_DOCSTRING_MULTIPLE_CONDITION_JOINING_PHRASES,
     ALL_DOCSTRING_NO_CONSUMER_CLAIM_PHRASES,
+    ALL_DOCSTRING_NO_INLINE_LITERAL_CLAIM_PHRASES,
+    ALL_DOCSTRING_NON_CONSTANT_REFERENCE_MARKERS,
+    ALL_GENERIC_CHECK_NAME_TOKENS,
+    ALL_NAMING_CONVENTION_DESCRIPTOR_TOKENS,
     DOCSTRING_FALLBACK_BRANCH_MINIMUM_ROUTE_COUNT,
+    DOCSTRING_REFERENCE_MARKER_WINDOW,
     DOCSTRING_TRIVIAL_FUNCTION_BODY_LINE_LIMIT,
     MAX_CLASS_DOCSTRING_PUBLIC_METHOD_ISSUES,
     MAX_DOCSTRING_ARGS_SIGNATURE_ISSUES,
     MAX_DOCSTRING_FALLBACK_BRANCH_ISSUES,
     MAX_DOCSTRING_FORMAT_ISSUES,
+    MAX_DOCSTRING_INLINE_LITERAL_CLAIM_ISSUES,
     MAX_DOCSTRING_NO_CONSUMER_CLAIM_ISSUES,
+    MAX_DOCSTRING_STEP_DISPATCH_ISSUES,
+    MAX_DOCSTRING_TUPLE_ENUMERATION_ISSUES,
+    MAX_DOCSTRING_UNDEFINED_CONSTANT_ISSUES,
+    MAX_MODULE_DOCSTRING_CHECK_ROSTER_ISSUES,
+    MINIMUM_NAMED_LINEAR_STEPS_FOR_DISPATCH_CHECK,
+    MINIMUM_PUBLIC_CHECKS_FOR_MODULE_DOCSTRING_ROSTER,
     MINIMUM_PUBLIC_METHODS_FOR_CLASS_DOCSTRING_BREADTH,
+    MINIMUM_TOKENS_FOR_DISPATCH_CALLEE,
+    MINIMUM_TUPLE_MEMBERS_FOR_DOCSTRING_ENUMERATION,
 )
 from hooks_constants.code_rules_enforcer_constants import (  # noqa: E402
+    ALL_CAPS_WITH_UNDERSCORE_PATTERN,
     ALL_DOCSTRING_ARGS_SECTION_HEADERS,
     ALL_DOCSTRING_TERMINATING_SECTION_HEADERS,
     ALL_SELF_AND_CLS_PARAMETER_NAMES,
     DOCSTRING_ARG_ENTRY_PATTERN,
+    IDENTIFIER_SHAPED_TUPLE_MEMBER_PATTERN,
+    INLINE_CODE_TOKEN_PATTERN,
 )
 
 
@@ -619,3 +639,599 @@ def check_docstring_no_consumer_claim(content: str, file_path: str) -> list[str]
         if len(issues) >= MAX_DOCSTRING_NO_CONSUMER_CLAIM_ISSUES:
             break
     return issues[:MAX_DOCSTRING_NO_CONSUMER_CLAIM_ISSUES]
+
+
+def _module_docstring_claims_no_inline_literal(module_docstring: str) -> str:
+    collapsed_docstring = " ".join(module_docstring.lower().split())
+    for each_phrase in ALL_DOCSTRING_NO_INLINE_LITERAL_CLAIM_PHRASES:
+        if each_phrase in collapsed_docstring:
+            return each_phrase
+    return ""
+
+
+def check_docstring_no_inline_literal_claim(content: str, file_path: str) -> list[str]:
+    """Flag a module docstring that asserts no literals appear inline elsewhere.
+
+    A constants-module docstring claiming "no literals appear inline in the
+    dispatcher" makes an unverifiable completeness claim about a companion file.
+    The claim drifts the moment a literal lands inline in that companion — a deny
+    or block reason left inline in the dispatcher contradicts the docstring even
+    though the constants file under edit never changed. This is the deterministic
+    slice of Category O6 (docstring prose vs implementation drift) and a
+    no-transitional-language violation in its own right: a docstring describes
+    what the module holds, not the absence of literals in a sibling. Rephrase to
+    state what the module centralizes, or drop the no-inline-literal sentence.
+
+    Args:
+        content: The source text to inspect.
+        file_path: The path the source will be written to, used for exemptions.
+
+    Returns:
+        One issue when the module docstring carries a no-inline-literal claim,
+        capped at the module limit.
+    """
+    if is_strict_test_file(file_path):
+        return []
+    try:
+        parsed_tree = ast.parse(content)
+    except SyntaxError:
+        return []
+    module_docstring = ast.get_docstring(parsed_tree) or ""
+    matched_phrase = _module_docstring_claims_no_inline_literal(module_docstring)
+    if not matched_phrase:
+        return []
+    issues = [
+        f"Line 1: module docstring claims '{matched_phrase}' about a companion file "
+        "— an unverifiable completeness claim that drifts the moment a literal lands "
+        "inline; state what the module centralizes instead (Category O6 docstring-vs-"
+        "implementation drift)"
+    ]
+    return issues[:MAX_DOCSTRING_INLINE_LITERAL_CLAIM_ISSUES]
+
+
+def _module_docstring_summary_is_single_paragraph(module_docstring: str) -> bool:
+    stripped_text = module_docstring.strip()
+    if not stripped_text:
+        return False
+    return "\n" not in stripped_text
+
+
+def _module_public_check_names(parsed_tree: ast.Module) -> list[str]:
+    deduplicated_names: dict[str, None] = {}
+    for each_statement in parsed_tree.body:
+        if not isinstance(each_statement, (ast.FunctionDef, ast.AsyncFunctionDef)):
+            continue
+        if not each_statement.name.startswith("check_"):
+            continue
+        if _function_is_private_or_dunder(each_statement.name):
+            continue
+        deduplicated_names[each_statement.name] = None
+    return list(deduplicated_names)
+
+
+def _distinctive_name_tokens(check_name: str) -> list[str]:
+    return [
+        each_token
+        for each_token in _name_tokens(check_name)
+        if each_token.lower() not in ALL_GENERIC_CHECK_NAME_TOKENS
+    ]
+
+
+def _docstring_mentions_check(docstring_text: str, check_name: str) -> bool:
+    lowered_docstring = docstring_text.lower()
+    if check_name.lower() in lowered_docstring:
+        return True
+    distinctive_tokens = _distinctive_name_tokens(check_name)
+    if not distinctive_tokens:
+        return True
+    return any(each_token.lower() in lowered_docstring for each_token in distinctive_tokens)
+
+
+def check_module_docstring_names_public_checks(content: str, file_path: str) -> list[str]:
+    """Flag a one-line module docstring that omits a public ``check_*`` function.
+
+    A check-registry module whose docstring is a single summary paragraph names
+    each check it dispatches, so a reader trusts that one line to be the full
+    roster. When the module grows a public ``check_*`` entry point the summary
+    never names, the enumeration under-describes the module — the
+    docstring-prose-vs-implementation drift the repo flags as Category O6/O8.
+    A check counts as named when the full ``check_*`` name, or any distinctive
+    (non-generic) underscore-separated token of it, appears in the summary;
+    generic tokens (``check``, ``test``, ``tests``) do not count. A module with
+    two or more public checks and any check the summary never names is reported
+    so the summary names the full roster. Modules with a multi-paragraph
+    docstring body are left to the audit lane, since their prose can carry the
+    roster without naming each check by name. This check covers hook
+    infrastructure, where the affected check registries live.
+
+    Args:
+        content: The source text to inspect.
+        file_path: The path the source will be written to, used for exemptions.
+
+    Returns:
+        One issue per public check the single-paragraph module docstring omits,
+        capped at the module limit.
+    """
+    if is_strict_test_file(file_path):
+        return []
+    try:
+        parsed_tree = ast.parse(content)
+    except SyntaxError:
+        return []
+    module_docstring = ast.get_docstring(parsed_tree) or ""
+    if not _module_docstring_summary_is_single_paragraph(module_docstring):
+        return []
+    public_check_names = _module_public_check_names(parsed_tree)
+    if len(public_check_names) < MINIMUM_PUBLIC_CHECKS_FOR_MODULE_DOCSTRING_ROSTER:
+        return []
+    issues: list[str] = []
+    for each_name in public_check_names:
+        if _docstring_mentions_check(module_docstring, each_name):
+            continue
+        issues.append(
+            f"Line 1: module docstring omits public check {each_name}() — name every "
+            "public check_* function the module exposes so the roster is complete "
+            "(Category O6/O8 docstring-vs-implementation drift)"
+        )
+        if len(issues) >= MAX_MODULE_DOCSTRING_CHECK_ROSTER_ISSUES:
+            break
+    return issues[:MAX_MODULE_DOCSTRING_CHECK_ROSTER_ISSUES]
+
+
+def _module_string_tuple_members(parsed_tree: ast.Module) -> dict[str, frozenset[str]]:
+    members_by_constant: dict[str, frozenset[str]] = {}
+    for each_statement in parsed_tree.body:
+        if not isinstance(each_statement, ast.Assign):
+            continue
+        if not isinstance(each_statement.value, ast.Tuple):
+            continue
+        literal_members: set[str] = set()
+        every_member_is_identifier_shaped = True
+        for each_element in each_statement.value.elts:
+            if (
+                isinstance(each_element, ast.Constant)
+                and isinstance(each_element.value, str)
+                and IDENTIFIER_SHAPED_TUPLE_MEMBER_PATTERN.match(each_element.value)
+            ):
+                literal_members.add(each_element.value.lstrip("."))
+                continue
+            every_member_is_identifier_shaped = False
+            break
+        if not every_member_is_identifier_shaped:
+            continue
+        if len(literal_members) < MINIMUM_TUPLE_MEMBERS_FOR_DOCSTRING_ENUMERATION:
+            continue
+        for each_target in each_statement.targets:
+            if isinstance(each_target, ast.Name):
+                members_by_constant[each_target.id] = frozenset(literal_members)
+    return members_by_constant
+
+
+def _names_referenced_in_function(
+    function_node: ast.FunctionDef | ast.AsyncFunctionDef,
+) -> set[str]:
+    return {
+        each_node.id
+        for each_node in ast.walk(function_node)
+        if isinstance(each_node, ast.Name)
+    }
+
+
+def _docstring_inline_code_tokens(docstring_text: str) -> set[str]:
+    tokens: set[str] = set()
+    for each_match in INLINE_CODE_TOKEN_PATTERN.finditer(docstring_text):
+        token = each_match.group(1).strip().lstrip(".")
+        if token:
+            tokens.add(token)
+    return tokens
+
+
+def check_docstring_tuple_enumeration_match(content: str, file_path: str) -> list[str]:
+    """Flag a docstring enumeration that drifts from a literal tuple it reads.
+
+    The drift this catches: a function reads a module-level tuple of literal
+    string members and its docstring enumerates inline-code tokens that name
+    some of those members, but the enumerated set and the tuple membership
+    differ. A token the docstring lists that the tuple lacks, or a tuple member
+    the docstring omits, misleads a reader who trusts the prose enumeration to
+    match the detection set — the deterministic slice of Category O6
+    docstring-prose-vs-implementation drift. The check binds only when the
+    docstring's inline-code tokens overlap the tuple membership, so a docstring
+    that names unrelated attributes is left alone. This check covers hook
+    infrastructure, where the affected detection tuples live.
+
+    Args:
+        content: The source text to inspect.
+        file_path: The path the source will be written to, used for exemptions.
+
+    Returns:
+        One issue per function whose docstring enumeration diverges from the
+        tuple it reads, capped at the module limit.
+    """
+    if is_strict_test_file(file_path):
+        return []
+    try:
+        parsed_tree = ast.parse(content)
+    except SyntaxError:
+        return []
+    members_by_constant = _module_string_tuple_members(parsed_tree)
+    if not members_by_constant:
+        return []
+    issues: list[str] = []
+    for each_node in _walk_skipping_type_checking_blocks(parsed_tree):
+        if not isinstance(each_node, (ast.FunctionDef, ast.AsyncFunctionDef)):
+            continue
+        docstring_text = _function_docstring_text(each_node)
+        if not docstring_text:
+            continue
+        docstring_tokens = _docstring_inline_code_tokens(docstring_text)
+        if not docstring_tokens:
+            continue
+        referenced_names = _names_referenced_in_function(each_node)
+        for each_constant_name in referenced_names & set(members_by_constant):
+            tuple_members = members_by_constant[each_constant_name]
+            if not (docstring_tokens & tuple_members):
+                continue
+            if docstring_tokens == tuple_members:
+                continue
+            docstring_only = sorted(docstring_tokens - tuple_members)
+            tuple_only = sorted(tuple_members - docstring_tokens)
+            issues.append(
+                f"Line {each_node.lineno}: {each_node.name}() docstring enumerates "
+                f"{sorted(docstring_tokens)} but {each_constant_name} holds "
+                f"{sorted(tuple_members)} — docstring-only: {docstring_only}, "
+                f"tuple-only: {tuple_only}; match the enumeration to the tuple "
+                "(Category O6 docstring-vs-implementation drift)"
+            )
+            if len(issues) >= MAX_DOCSTRING_TUPLE_ENUMERATION_ISSUES:
+                return issues[:MAX_DOCSTRING_TUPLE_ENUMERATION_ISSUES]
+    return issues[:MAX_DOCSTRING_TUPLE_ENUMERATION_ISSUES]
+
+
+def _called_callee_name(statement: ast.stmt) -> str:
+    candidate_expression: ast.expr | None = None
+    if isinstance(statement, ast.Expr):
+        candidate_expression = statement.value
+    elif isinstance(statement, ast.Assign):
+        candidate_expression = statement.value
+    elif isinstance(statement, ast.AnnAssign):
+        candidate_expression = statement.value
+    if candidate_expression is None:
+        return ""
+    if isinstance(candidate_expression, ast.Await):
+        candidate_expression = candidate_expression.value
+    if not isinstance(candidate_expression, ast.Call):
+        return ""
+    return _call_callee_name(candidate_expression)
+
+
+def _called_callees_in_expression(expression: ast.expr) -> set[str]:
+    callees: set[str] = set()
+    for each_descendant in ast.walk(expression):
+        if not isinstance(each_descendant, ast.Call):
+            continue
+        callee_name = _call_callee_name(each_descendant)
+        if callee_name:
+            callees.add(callee_name)
+    return callees
+
+
+def _linear_step_callees(
+    function_node: ast.FunctionDef | ast.AsyncFunctionDef,
+) -> set[str]:
+    callees: set[str] = set()
+    for each_statement in function_node.body:
+        callee_name = _called_callee_name(each_statement)
+        if callee_name:
+            callees.add(callee_name)
+        if isinstance(each_statement, ast.If):
+            callees |= _called_callees_in_expression(each_statement.test)
+    return callees
+
+
+def _branch_guarded_dispatch_callees(
+    function_node: ast.FunctionDef | ast.AsyncFunctionDef,
+) -> set[str]:
+    callees: set[str] = set()
+    for each_statement in function_node.body:
+        if not isinstance(each_statement, ast.If):
+            continue
+        for each_branch_statement in each_statement.body + each_statement.orelse:
+            for each_descendant in ast.walk(each_branch_statement):
+                if not isinstance(each_descendant, ast.If):
+                    continue
+                callees |= _called_callees_in_expression(each_descendant.test)
+    return callees
+
+
+def _bare_callee_name(qualified_callee: str) -> str:
+    return qualified_callee.rsplit(".", 1)[-1]
+
+
+def _docstring_names_all_callee_tokens(docstring_text: str, callee_name: str) -> bool:
+    bare_name = _bare_callee_name(callee_name)
+    lowered_docstring = docstring_text.lower()
+    if bare_name.lower() in lowered_docstring:
+        return True
+    callee_tokens = _name_tokens(bare_name)
+    if not callee_tokens:
+        return False
+    return all(each_token.lower() in lowered_docstring for each_token in callee_tokens)
+
+
+def check_docstring_step_enumeration_dispatch_coverage(
+    content: str, file_path: str
+) -> list[str]:
+    """Flag a step-enumeration docstring that omits a conditional dispatch call.
+
+    The drift this catches: a function whose docstring enumerates a linear
+    sequence of steps (``Navigates ..., searches ..., clicks ..., uploads ...``)
+    matching the body's linear-step calls, while the body also routes to a
+    corrective workflow step inside an ``if``/``elif`` branch — a cancel-and-reinitiate
+    or replace-target-row step — whose name the prose never spells out. A reader
+    who trusts the step list to be complete misses that the function can take that
+    conditional path. This is the deterministic slice of Category O4 (step-ordering
+    narrative): a body that guards a branch-only workflow step the enumeration omits.
+
+    A linear-step call is one made as a top-level statement or inside the ``If.test``
+    guard of a top-level ``if`` (``if not await self.navigate(): return``). A
+    dispatch step is a call inside a guard (``If.test``) nested within an
+    ``if``/``elif`` branch (``if not await cancel_and_reinitiate_update(...): return``)
+    that is never also a linear step — the same control-flow-gating shape as a
+    linear step, so plain (unguarded) logging, screenshot, or method-on-local
+    calls inside a branch body are not dispatch steps. The check binds only when
+    the docstring already names two or more linear-step callees by their
+    underscore tokens, proving the prose is a step enumeration describing this
+    body. A dispatch-step callee with two or more underscore tokens, none of
+    whose tokens appear in the prose, is flagged.
+
+    Args:
+        content: The source text to inspect.
+        file_path: The path the source will be written to, used for exemptions.
+
+    Returns:
+        One issue per conditional dispatch call the step enumeration omits, capped
+        at the module limit.
+    """
+    if is_test_file(file_path) or is_hook_infrastructure(file_path):
+        return []
+    try:
+        parsed_tree = ast.parse(content)
+    except SyntaxError:
+        return []
+    issues: list[str] = []
+    for each_node in _walk_skipping_type_checking_blocks(parsed_tree):
+        if not isinstance(each_node, (ast.FunctionDef, ast.AsyncFunctionDef)):
+            continue
+        if _function_has_exempt_decorator(each_node):
+            continue
+        docstring_text = _function_docstring_text(each_node)
+        if not docstring_text:
+            continue
+        linear_step_callees = _linear_step_callees(each_node)
+        named_linear_steps = [
+            each_callee
+            for each_callee in linear_step_callees
+            if _docstring_names_all_callee_tokens(docstring_text, each_callee)
+        ]
+        if len(named_linear_steps) < MINIMUM_NAMED_LINEAR_STEPS_FOR_DISPATCH_CHECK:
+            continue
+        branch_only_callees = (
+            _branch_guarded_dispatch_callees(each_node) - linear_step_callees
+        )
+        for each_callee in sorted(branch_only_callees):
+            if len(_name_tokens(_bare_callee_name(each_callee))) < MINIMUM_TOKENS_FOR_DISPATCH_CALLEE:
+                continue
+            if _docstring_names_all_callee_tokens(docstring_text, each_callee):
+                continue
+            issues.append(
+                f"Line {each_node.lineno}: {each_node.name}() docstring enumerates linear "
+                f"steps but omits the conditional dispatch step {each_callee}() the body "
+                "guards inside a branch — add the corrective-path step to the enumeration "
+                "(Category O4 step-ordering narrative drift)"
+            )
+            if len(issues) >= MAX_DOCSTRING_STEP_DISPATCH_ISSUES:
+                return issues[:MAX_DOCSTRING_STEP_DISPATCH_ISSUES]
+    return issues[:MAX_DOCSTRING_STEP_DISPATCH_ISSUES]
+
+
+def _imported_binding_names(import_node: ast.Import | ast.ImportFrom) -> set[str]:
+    bound_names: set[str] = set()
+    for each_alias in import_node.names:
+        bound_names.add(each_alias.asname or each_alias.name.split(".", 1)[0])
+    return bound_names
+
+
+def _module_defined_and_imported_names(parsed_tree: ast.Module) -> set[str]:
+    defined_names: set[str] = set()
+    for each_node in ast.walk(parsed_tree):
+        if isinstance(each_node, (ast.Import, ast.ImportFrom)):
+            defined_names |= _imported_binding_names(each_node)
+        elif isinstance(each_node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)):
+            defined_names.add(each_node.name)
+        elif isinstance(each_node, ast.Name) and isinstance(each_node.ctx, ast.Store):
+            defined_names.add(each_node.id)
+    return defined_names
+
+
+def _module_attribute_access_names(parsed_tree: ast.Module) -> set[str]:
+    attribute_names: set[str] = set()
+    for each_node in ast.walk(parsed_tree):
+        if isinstance(each_node, ast.Attribute):
+            attribute_names.add(each_node.attr)
+    return attribute_names
+
+
+def _docstring_constant_node_ids(parsed_tree: ast.Module) -> set[int]:
+    docstring_node_ids: set[int] = set()
+    for each_node in ast.walk(parsed_tree):
+        if not isinstance(
+            each_node, (ast.Module, ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)
+        ):
+            continue
+        body_statements = each_node.body
+        if not body_statements or not _statement_is_docstring(body_statements[0]):
+            continue
+        first_statement = body_statements[0]
+        assert isinstance(first_statement, ast.Expr)
+        docstring_node_ids.add(id(first_statement.value))
+    return docstring_node_ids
+
+
+def _module_string_literal_word_runs(parsed_tree: ast.Module) -> set[str]:
+    docstring_node_ids = _docstring_constant_node_ids(parsed_tree)
+    word_runs: set[str] = set()
+    for each_node in ast.walk(parsed_tree):
+        if not (isinstance(each_node, ast.Constant) and isinstance(each_node.value, str)):
+            continue
+        if id(each_node) in docstring_node_ids:
+            continue
+        for each_run in re.findall(r"[A-Za-z0-9_]+", each_node.value):
+            if ALL_CAPS_WITH_UNDERSCORE_PATTERN.match(each_run):
+                word_runs.add(each_run)
+    return word_runs
+
+
+def _name_word_prefix_families(all_supporting_names: set[str]) -> set[str]:
+    prefix_families: set[str] = set()
+    for each_name in all_supporting_names:
+        leading_word = each_name.split("_", 1)[0]
+        prefix_families.add(leading_word)
+    return prefix_families
+
+
+def _token_is_word_run_of_any_name(token: str, all_supporting_names: set[str]) -> bool:
+    return any(f"_{token}_" in f"_{each_name}_" for each_name in all_supporting_names)
+
+
+def _docstring_words(docstring_text: str) -> list[str]:
+    return [
+        each_word.strip(".,:;()[]{}'\"`")
+        for each_word in docstring_text.replace("`", " ").split()
+    ]
+
+
+def _docstring_frames_token_as_non_constant_reference(
+    token: str, docstring_text: str
+) -> bool:
+    if any(
+        f"{token}{each_suffix}" in docstring_text
+        for each_suffix in ALL_DOCSTRING_FILE_REFERENCE_SUFFIXES
+    ):
+        return True
+    words = _docstring_words(docstring_text)
+    for each_index, each_word in enumerate(words):
+        if each_word != token:
+            continue
+        neighbors = words[max(each_index - DOCSTRING_REFERENCE_MARKER_WINDOW, 0) : each_index + DOCSTRING_REFERENCE_MARKER_WINDOW + 1]
+        if any(
+            each_neighbor.lower() in ALL_DOCSTRING_NON_CONSTANT_REFERENCE_MARKERS
+            for each_neighbor in neighbors
+        ):
+            return True
+    return False
+
+
+def _docstring_constant_token_is_supported(
+    token: str, parsed_tree: ast.Module, all_known_names: set[str], docstring_text: str
+) -> bool:
+    supporting_predicates = (
+        lambda: token in all_known_names,
+        lambda: token in ALL_NAMING_CONVENTION_DESCRIPTOR_TOKENS,
+        lambda: token in _module_attribute_access_names(parsed_tree),
+        lambda: token in _module_string_literal_word_runs(parsed_tree),
+        lambda: _token_is_word_run_of_any_name(token, all_known_names),
+        lambda: _docstring_frames_token_as_non_constant_reference(token, docstring_text),
+        lambda: token.split("_", 1)[0] in _name_word_prefix_families(all_known_names),
+    )
+    return any(each_predicate() for each_predicate in supporting_predicates)
+
+
+def _docstring_constant_tokens(docstring_text: str) -> set[str]:
+    candidate_tokens: set[str] = set()
+    for each_word in docstring_text.replace("`", " ").split():
+        stripped_word = each_word.strip(".,:;()[]{}'\"")
+        if stripped_word.startswith("__") and stripped_word.endswith("__"):
+            continue
+        if ALL_CAPS_WITH_UNDERSCORE_PATTERN.match(stripped_word):
+            candidate_tokens.add(stripped_word)
+    return candidate_tokens
+
+
+def _documentable_nodes_with_docstrings(
+    parsed_tree: ast.Module,
+) -> list[tuple[int, str]]:
+    documentable: list[tuple[int, str]] = []
+    module_docstring = ast.get_docstring(parsed_tree)
+    if module_docstring:
+        documentable.append((1, module_docstring))
+    for each_node in _walk_skipping_type_checking_blocks(parsed_tree):
+        if not isinstance(
+            each_node, (ast.FunctionDef, ast.AsyncFunctionDef, ast.ClassDef)
+        ):
+            continue
+        node_docstring = ast.get_docstring(each_node)
+        if node_docstring:
+            documentable.append((each_node.lineno, node_docstring))
+    return documentable
+
+
+def check_docstring_names_undefined_constant(content: str, file_path: str) -> list[str]:
+    """Flag a docstring naming an UPPER_SNAKE constant nothing in the module backs.
+
+    The drift this catches: a docstring names an all-caps, underscore-joined
+    token as a contract identifier (``NATIVE_EVALUATE_FUNCTION_NAME``) while the
+    enclosing module carries no supporting reference for it. A reader who trusts
+    the docstring to name a real symbol finds nothing — the deterministic slice
+    of Category O6 docstring-prose-vs-implementation drift where the named token
+    is structurally a constant and unresolvable against the module.
+
+    A token counts as supported, and is left alone, when any of these holds: it
+    is defined at module scope or imported; it is a naming-convention descriptor
+    (``UPPER_SNAKE_CASE`` and its siblings, describing a style, not a symbol); it
+    is the attribute of an attribute access in the body (``os.O_NOFOLLOW``,
+    ``config.timing.MAX_DELAY``, resolving stdlib and dotted-import constants);
+    it is an all-caps word run inside a string literal (an env-var key read via
+    ``os.environ[...]`` or ``os.getenv(...)``, an API enum string value, a doc
+    stem in ``CODE_RULES.md``); it is a contiguous word run of a defined or
+    imported name (``GH_TOKEN`` within ``ALL_GH_TOKEN_ENV_VAR_NAMES``); it shares
+    a leading word component with a defined or imported name, marking the same
+    enum family (``MODE_CLASSIFY`` beside an imported ``MODE_STRICT``); or the
+    docstring prose frames it as a non-constant reference — followed by a file
+    suffix (``CODE_RULES.md``) or sitting within two words of a marker such as
+    ``rule``, ``doc``, ``file``, ``env``, ``variable``, ``set``, ``read``,
+    ``per``, ``follows``, or ``see`` (``per CODE_RULES``, ``LLM_SETTINGS_ROOT is
+    set to``). Single-segment all-caps acronyms (``HTTP``, ``JSON``) and dunder
+    names (``__all__``) are not constants and are left alone.
+
+    Args:
+        content: The source text to inspect.
+        file_path: The path the source will be written to, used for exemptions.
+
+    Returns:
+        One issue per docstring token that no module reference backs, capped at
+        the module limit.
+    """
+    if is_test_file(file_path) or is_hook_infrastructure(file_path):
+        return []
+    try:
+        parsed_tree = ast.parse(content)
+    except SyntaxError:
+        return []
+    known_names = _module_defined_and_imported_names(parsed_tree)
+    issues: list[str] = []
+    for each_line_number, each_docstring in _documentable_nodes_with_docstrings(parsed_tree):
+        for each_token in sorted(_docstring_constant_tokens(each_docstring)):
+            if _docstring_constant_token_is_supported(
+                each_token, parsed_tree, known_names, each_docstring
+            ):
+                continue
+            issues.append(
+                f"Line {each_line_number}: docstring names '{each_token}' which the "
+                "module neither defines at module scope nor imports — name the real "
+                "symbol or drop the reference (Category O6 docstring-vs-implementation "
+                "drift)"
+            )
+            if len(issues) >= MAX_DOCSTRING_UNDEFINED_CONSTANT_ISSUES:
+                return issues[:MAX_DOCSTRING_UNDEFINED_CONSTANT_ISSUES]
+    return issues[:MAX_DOCSTRING_UNDEFINED_CONSTANT_ISSUES]

package/hooks/blocking/code_rules_enforcer.py

@@ -69,7 +69,12 @@ from code_rules_docstrings import (  # noqa: E402
     check_docstring_args_match_signature,
     check_docstring_fallback_branch_coverage,
     check_docstring_format,
+    check_docstring_names_undefined_constant,
     check_docstring_no_consumer_claim,
+    check_docstring_no_inline_literal_claim,
+    check_docstring_step_enumeration_dispatch_coverage,
+    check_docstring_tuple_enumeration_match,
+    check_module_docstring_names_public_checks,
 )
 from code_rules_duplicate_body import (  # noqa: E402
     advise_cross_skill_duplicate_helper,
@@ -255,9 +260,26 @@ def validate_content(
         all_issues.extend(check_docstring_args_match_signature(effective_content, file_path))
         all_issues.extend(check_docstring_fallback_branch_coverage(effective_content, file_path))
         all_issues.extend(check_docstring_no_consumer_claim(effective_content, file_path))
+        all_issues.extend(
+            check_docstring_no_inline_literal_claim(effective_content, file_path)
+        )
         all_issues.extend(
             check_class_docstring_names_public_methods(effective_content, file_path)
         )
+        all_issues.extend(
+            check_module_docstring_names_public_checks(effective_content, file_path)
+        )
+        all_issues.extend(
+            check_docstring_tuple_enumeration_match(effective_content, file_path)
+        )
+        all_issues.extend(
+            check_docstring_step_enumeration_dispatch_coverage(
+                effective_content, file_path
+            )
+        )
+        all_issues.extend(
+            check_docstring_names_undefined_constant(effective_content, file_path)
+        )
         all_issues.extend(
             check_boolean_naming(
                 effective_content,