fix(terraform): redact secrets nested inside list values, not just maps (#3644 follow-up)

graphify/extractors/terraform.py

@@ -35,12 +35,26 @@
 
 
 def _redact_value(key: str, value: object) -> object:
-    """Redact a sensitive attribute value; recurse into map values so a nested
-    `password` inside a `tags`/`connection` map is redacted too."""
+    """Redact a sensitive attribute value; recurse into map AND list values so a
+    nested `password` inside a `tags`/`connection` map — or inside a list of
+    objects — is redacted too.
+
+    HCL routinely nests objects inside tuples (`list(object(...))` variables,
+    `dynamic` blocks, tuple defaults), and `_parse_attr_value` turns those into
+    Python lists of dicts. Recursing into dicts but not lists left
+    `configs = [{ password = "x" }]` leaking verbatim while the map form
+    `config = { password = "x" }` was redacted — the value still reaches
+    graph.json and the MCP query/get_node surface unsanitized (#3644 follow-up).
+    List elements are recursed under the same key: the list branch is only
+    reached when `key` is NOT itself sensitive (a sensitive key redacts the whole
+    value above), so a scalar element carries no key signal and is returned
+    as-is, while a dict element is checked against its own inner keys."""
     if _SENSITIVE_KEY_RE.search(key):
         return _REDACTED
     if isinstance(value, dict):
         return {k: _redact_value(str(k), v) for k, v in value.items()}
+    if isinstance(value, (list, tuple)):
+        return [_redact_value(key, item) for item in value]
     return value
 
 

tests/test_terraform.py

@@ -359,3 +359,44 @@ def test_terraform_sensitive_attribute_values_are_redacted(tmp_path):
     assert "hunter2-super-secret" not in _json.dumps(node)
     assert "AKIAWHATEVER" not in _json.dumps(node)
     assert "leaky" not in _json.dumps(node)
+
+
+def test_terraform_redact_value_recurses_into_lists():
+    """A secret nested inside a list — a list of objects, or a nested list — must
+    be redacted, matching the map case. HCL routinely nests objects in tuples
+    (`list(object(...))`, dynamic blocks), and recursing into dicts but not lists
+    left those values leaking (#3644 follow-up)."""
+    from graphify.extractors.terraform import _redact_value
+
+    # list of objects: the secret-named inner key is redacted, ordinary key kept
+    assert _redact_value("connections", [{"host": "db", "password": "x"}]) == [
+        {"host": "db", "password": "[redacted]"}
+    ]
+    # nested list -> list -> object
+    assert _redact_value("stages", [[{"api_key": "sk-1"}]]) == [[{"api_key": "[redacted]"}]]
+    # a sensitive KEY still redacts the whole list value
+    assert _redact_value("passwords", ["a", "b"]) == "[redacted]"
+    # a non-secret list of scalars is left intact
+    assert _redact_value("ports", [80, 443]) == [80, 443]
+
+
+def test_terraform_secret_in_list_of_objects_is_redacted(tmp_path):
+    """End-to-end: a `password` inside a tuple-of-objects attribute must not reach
+    the graph verbatim, exactly as it wouldn't inside a map (#3644 follow-up)."""
+    body = """\
+resource "aws_x" "y" {
+  name = "app"
+  connections = [
+    { host = "db1", password = "leaky-in-list" },
+    { host = "db2", token = "tok-in-list" },
+  ]
+}
+"""
+    r = extract_terraform(_write(tmp_path, "conns.tf", body))
+    node = next(n for n in r["nodes"] if n["label"] == "aws_x.y")
+    conns = node["attributes"]["connections"]
+    assert conns[0]["host"] == "db1" and conns[0]["password"] == "[redacted]"
+    assert conns[1]["host"] == "db2" and conns[1]["token"] == "[redacted]"
+    import json as _json
+    assert "leaky-in-list" not in _json.dumps(node)
+    assert "tok-in-list" not in _json.dumps(node)