fix(client): explain missing Google signatures after gateway rejection

Append bounded recovery guidance only to HTTP 400 responses reporting a missing thought_signature. Both Chat request paths use this sanitizer; generic gateway preflight remains unchanged. Add release notes for this repair, Markdown clipboard behavior and the embedded CU JPEG port.

Validation: hermetic llm_client and Google signature nextest selection: 41 tests run, 41 passed, 12357 skipped. TUI library cargo check passed. Version check and documentation check (23 topics) passed; git diff --check passed. Root npm gate passed on the gate-repair slice: wrapper 66, SDK 9, web 446 tests; final-head full sweep follows.

CHANGELOG.md

@@ -7,7 +7,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
-## [0.9.13] - 2026-09-10
+## [0.9.13] - 2026-09-11
 
 Codewhale v0.9.13 source candidate addresses integrity issues in 0.9.12:
 multiline paste is one paste again, truncated tool arguments can no longer execute, strict
@@ -21,6 +21,16 @@ reconnect.
 
 ### Fixed
 
+- Direct composer paste (`Ctrl-V`) preserves rich clipboard headings, lists,
+  links, tables and code as Markdown, with plain-text fallback. Whole-answer
+  copy preserves authored Markdown without terminal wrapping; settings fields
+  still paste literal text.
+- The embedded Computer Use bundle includes the marketplace's JPEG capture
+  and bounded-payload fixes, so new built-in installations receive them too.
+- A provider's missing-`thought_signature` HTTP 400 now explains how to recover:
+  use the built-in Google provider and start a new session, or verify the
+  gateway's signature handling. Working gateways are not blocked by a broader
+  preflight check (#6048).
 - The sandbox-elevation prompt shows every option, `Abort` included. The card
   was a fixed 22 rows centred on the frame; inside its border and padding that
   left at most 18 usable rows against 20 to 23 rows of content, with no scroll

crates/tui/src/llm_client/mod.rs

@@ -653,7 +653,7 @@ pub(crate) fn sanitize_http_error_body(
         if let Some(code) = explicit_quota_code(body) {
             return format!("{message} (provider error code: {code})");
         }
-        return message;
+        return with_thought_signature_recovery(status, message);
     }
 
     if is_probably_html(body) {
@@ -696,7 +696,24 @@ pub(crate) fn sanitize_http_error_body(
         return format!("{provider} API returned an HTML error page (HTTP {status}): {text}");
     }
 
-    truncate_for_error(&collapse_whitespace(body), 2_000)
+    with_thought_signature_recovery(
+        status,
+        truncate_for_error(&collapse_whitespace(body), 2_000),
+    )
+}
+
+/// A gateway may manage signatures itself, so only explain an actual rejection
+/// instead of preemptively blocking every Gemini-compatible route.
+fn with_thought_signature_recovery(status: u16, mut message: String) -> String {
+    let lower = message.to_ascii_lowercase();
+    if status == 400 && lower.contains("missing") && lower.contains("thought_signature") {
+        const HINT: &str = " Use the built-in `google` provider for Gemini and start a new session; \
+                           earlier tool calls cannot recover missing thought signatures. If you use a \
+                           gateway, confirm that it preserves or manages Google thought signatures.";
+        message = truncate_for_error(&message, 2_000 - HINT.len());
+        message.push_str(HINT);
+    }
+    message
 }
 
 fn looks_like_authentication_failure(body: &str) -> bool {

crates/tui/src/llm_client/tests.rs

@@ -150,6 +150,34 @@ fn generic_429_stays_rate_limited_and_retryable() {
     assert!(error.is_retryable());
 }
 
+#[test]
+fn missing_google_signature_400_explains_recovery_without_widening_gateway_preflight() {
+    let message = "Function call is missing a thought_signature in functionCall parts";
+    for body in [
+        message.to_string(),
+        serde_json::json!({"error": {"message": message}}).to_string(),
+        format!("{message} {}", "详情".repeat(2_000)),
+    ] {
+        let safe = sanitize_http_error_body(Some("OpenAI-compatible"), 400, &body);
+        assert!(safe.starts_with(message));
+        assert!(safe.contains("built-in `google` provider"));
+        assert!(safe.contains("start a new session"));
+        assert!(safe.contains("gateway"));
+        assert!(
+            safe.chars().count() <= 2_003,
+            "error bound survives the hint"
+        );
+    }
+    for (status, body) in [
+        (429, message),
+        (200, message),
+        (400, "Invalid thought_signature"),
+        (400, "Missing required parameter: model"),
+    ] {
+        assert_eq!(sanitize_http_error_body(None, status, body), body);
+    }
+}
+
 #[tokio::test]
 async fn retry_loop_stops_after_one_typed_quota_failure() {
     let mut calls = 0;