diff --git a/pyproject.toml b/pyproject.toml
index 232892d..affb80b 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "goodeye"
-version = "0.25.3"
+version = "0.25.4"
 description = "Goodeye CLI: a private home for the skills your AI follows and the verifiers its work must pass, from the terminal."
 readme = "README.md"
 license = { file = "LICENSE" }
diff --git a/src/goodeye_cli/wire.py b/src/goodeye_cli/wire.py
index 472faa5..614592b 100644
--- a/src/goodeye_cli/wire.py
+++ b/src/goodeye_cli/wire.py
@@ -8,9 +8,29 @@
 from __future__ import annotations
 
 from datetime import datetime
-from typing import Any
+from typing import Annotated, Any
 
-from pydantic import BaseModel, ConfigDict, Field
+from pydantic import BaseModel, BeforeValidator, ConfigDict, Field
+
+
+def _empty_when_null(value: Any) -> Any:
+    """Coerce an explicit JSON null to an empty string.
+
+    Optional server columns arrive as null once a caller clears them, rather
+    than being omitted from the payload. Every field annotated with
+    ``NullableStr`` already defaults to "", so an absent value is expected;
+    without this coercion one null row fails validation and takes down the
+    whole response, including the rows that parsed fine.
+    """
+    return "" if value is None else value
+
+
+NullableStr = Annotated[str, BeforeValidator(_empty_when_null)]
+"""A string field that tolerates null on the wire and reads as "" in Python.
+
+Use for any optional text the server may return as null. Consumers keep a
+plain ``str``, so callers that format or concatenate the value stay safe.
+"""
 
 
 class _WireBase(BaseModel):
@@ -108,8 +128,8 @@ class WorkflowSummary(_WireBase):
     id: str
     name: str
     current_version: int
-    description: str = ""
-    outcome: str = ""
+    description: NullableStr = ""
+    outcome: NullableStr = ""
     tags: list[str] = Field(default_factory=list)
     updated_at: datetime | None = None
     owner_user_id: str | None = None
@@ -136,8 +156,8 @@ class WorkflowSearchItem(_WireBase):
     match_reason: str
     slug: str | None = None
     name: str | None = None
-    description: str = ""
-    outcome: str = ""
+    description: NullableStr = ""
+    outcome: NullableStr = ""
     tags: list[str] = Field(default_factory=list)
 
 
@@ -207,8 +227,8 @@ class WorkflowDetail(_WireBase):
     name: str
     version: int
     body: str
-    description: str = ""
-    outcome: str = ""
+    description: NullableStr = ""
+    outcome: NullableStr = ""
     tags: list[str] = Field(default_factory=list)
     owner_user_id: str | None = None
     updated_at: datetime | None = None
@@ -255,7 +275,7 @@ class WorkflowFilePatchResult(_WireBase):
     version: int
     version_token: str
     name: str
-    slug: str = ""
+    slug: NullableStr = ""
     changed: list[str] = Field(default_factory=list)
     deleted: list[str] = Field(default_factory=list)
     carried_forward: int = 0
@@ -369,7 +389,7 @@ class DesignCheckCriterion(_WireBase):
 
     criterion: str
     passed: bool
-    reason: str = ""
+    reason: NullableStr = ""
 
 
 class DesignChecks(_WireBase):
@@ -395,8 +415,8 @@ class TemplateSummary(_WireBase):
     handle: str
     owner_user_id: str
     latest_version: int
-    description: str = ""
-    outcome: str = ""
+    description: NullableStr = ""
+    outcome: NullableStr = ""
     tags: list[str] = Field(default_factory=list)
     publishing_handle: str
     safety_verification_status: str = "unverified"
@@ -415,11 +435,11 @@ class TemplateSearchItem(_WireBase):
     id: str
     rank: int
     match_reason: str
-    slug: str = ""
-    name: str = ""
-    handle: str = ""
-    description: str = ""
-    outcome: str = ""
+    slug: NullableStr = ""
+    name: NullableStr = ""
+    handle: NullableStr = ""
+    description: NullableStr = ""
+    outcome: NullableStr = ""
     tags: list[str] = Field(default_factory=list)
 
 
@@ -486,8 +506,8 @@ class TemplateDetail(_WireBase):
     owner_user_id: str
     version: int
     body: str
-    description: str = ""
-    outcome: str = ""
+    description: NullableStr = ""
+    outcome: NullableStr = ""
     tags: list[str] = Field(default_factory=list)
     release_notes: str | None = None
     publishing_handle: str
@@ -797,7 +817,7 @@ class ImageGeneratorSummary(_WireBase):
 
     generator_id: str
     name: str
-    description: str = ""
+    description: NullableStr = ""
     current_version: int
     version_token: str
     status: str
@@ -816,7 +836,7 @@ class ImageGeneratorDetail(_WireBase):
 
     generator_id: str
     name: str
-    description: str = ""
+    description: NullableStr = ""
     current_version: int
     version: int
     version_token: str
@@ -836,7 +856,7 @@ class ImageGeneratorDeployResult(_WireBase):
 
     generator_id: str
     name: str
-    description: str = ""
+    description: NullableStr = ""
     current_version: int
     version: int
     version_token: str
diff --git a/tests/test_wire_nullable_strings.py b/tests/test_wire_nullable_strings.py
new file mode 100644
index 0000000..c4ba74c
--- /dev/null
+++ b/tests/test_wire_nullable_strings.py
@@ -0,0 +1,121 @@
+"""Tests that optional wire strings tolerate an explicit JSON null.
+
+Optional server columns come back as null once a caller clears them. These
+fields already default to "", so a null must read the same way rather than
+failing validation and taking the whole response down with it.
+"""
+
+from __future__ import annotations
+
+import pytest
+from pydantic import ValidationError
+
+from goodeye_cli.wire import (
+    TemplateList,
+    TemplateSearchResponse,
+    TemplateSummary,
+    WorkflowDetail,
+    WorkflowList,
+    WorkflowSearchResponse,
+    WorkflowSummary,
+)
+
+
+def _template_row(**overrides: object) -> dict[str, object]:
+    row: dict[str, object] = {
+        "id": "tmpl-1",
+        "slug": "writing-humanizer",
+        "name": "writing-humanizer",
+        "handle": "example",
+        "owner_user_id": "user-1",
+        "latest_version": 10,
+        "publishing_handle": "example",
+    }
+    row.update(overrides)
+    return row
+
+
+def _workflow_row(**overrides: object) -> dict[str, object]:
+    row: dict[str, object] = {"id": "skill-1", "name": "example-skill", "current_version": 1}
+    row.update(overrides)
+    return row
+
+
+@pytest.mark.parametrize("field", ["outcome", "description"])
+def test_template_summary_accepts_null(field: str) -> None:
+    summary = TemplateSummary.model_validate(_template_row(**{field: None}))
+    assert getattr(summary, field) == ""
+
+
+@pytest.mark.parametrize("field", ["outcome", "description"])
+def test_workflow_summary_accepts_null(field: str) -> None:
+    summary = WorkflowSummary.model_validate(_workflow_row(**{field: None}))
+    assert getattr(summary, field) == ""
+
+
+def test_null_outcome_does_not_fail_the_whole_template_list() -> None:
+    """One cleared row must not take down the rows that parsed fine."""
+    payload = {
+        "items": [_template_row(id="tmpl-1"), _template_row(id="tmpl-2", outcome=None)],
+        "next_cursor": None,
+    }
+    listing = TemplateList.model_validate(payload)
+    assert [item.id for item in listing.items] == ["tmpl-1", "tmpl-2"]
+    assert listing.items[1].outcome == ""
+
+
+def test_null_outcome_does_not_fail_the_whole_workflow_list() -> None:
+    payload = {"items": [_workflow_row(), _workflow_row(id="skill-2", outcome=None)]}
+    listing = WorkflowList.model_validate(payload)
+    assert listing.items[1].outcome == ""
+
+
+def test_template_search_accepts_null_strings() -> None:
+    payload = {
+        "items": [
+            {
+                "id": "tmpl-1",
+                "rank": 1,
+                "match_reason": "name match",
+                "slug": None,
+                "name": None,
+                "handle": None,
+                "description": None,
+                "outcome": None,
+            }
+        ],
+        "query": "humanize",
+        "limit": 10,
+    }
+    item = TemplateSearchResponse.model_validate(payload).items[0]
+    assert (item.slug, item.name, item.handle, item.description, item.outcome) == ("",) * 5
+
+
+def test_workflow_search_accepts_null_strings() -> None:
+    payload = {
+        "items": [{"id": "skill-1", "rank": 1, "match_reason": "m", "description": None}],
+        "query": "humanize",
+        "limit": 10,
+    }
+    assert WorkflowSearchResponse.model_validate(payload).items[0].description == ""
+
+
+def test_workflow_detail_accepts_null_outcome() -> None:
+    detail = WorkflowDetail.model_validate(
+        {"id": "skill-1", "name": "n", "version": 1, "body": "b", "outcome": None}
+    )
+    assert detail.outcome == ""
+
+
+def test_absent_and_present_values_are_unchanged() -> None:
+    """The coercion must only touch null, leaving normal payloads alone."""
+    absent = TemplateSummary.model_validate(_template_row())
+    assert absent.outcome == ""
+    present = TemplateSummary.model_validate(_template_row(outcome="Ship faster."))
+    assert present.outcome == "Ship faster."
+
+
+def test_wrong_types_are_still_rejected() -> None:
+    """Tolerating null must not turn the field into an anything-goes field."""
+    with pytest.raises(ValidationError):
+        TemplateSummary.model_validate(_template_row(outcome=123))
diff --git a/uv.lock b/uv.lock
index 8c0e996..9ea194a 100644
--- a/uv.lock
+++ b/uv.lock
@@ -176,7 +176,7 @@ wheels = [
 
 [[package]]
 name = "goodeye"
-version = "0.25.3"
+version = "0.25.4"
 source = { editable = "." }
 dependencies = [
     { name = "httpx" },