Normalize Codex Responses input shape

Ensure the Codex OAuth Responses proxy forwards list-shaped input for empty continuation turns while preserving string prompts as user input items. Add regression coverage for the gpt-5.5 continuation case and document the proxy contract.

plugins/_oauth/AGENTS.md

@@ -40,7 +40,7 @@
 - Stored upstream base URLs and OAuth token endpoints must be validated against provider-owned allowlists before sending bearer or refresh tokens.
 - Browser callback providers must support manual callback paste when the browser cannot reach the local callback route.
 - Local proxy routes must remain loopback or token protected and must not add broad CORS access.
-- Codex Responses proxy requests must include Codex client metadata and compatibility headers such as `client_metadata`, `x-codex-installation-id`, `originator`, `session-id`, and `thread-id`.
+- Codex Responses proxy requests must include Codex client metadata and compatibility headers such as `client_metadata`, `x-codex-installation-id`, `originator`, `session-id`, and `thread-id`, and must forward `input` as a list for upstream Codex compatibility.
 
 ## Work Guidance
 

plugins/_oauth/helpers/codex.py

@@ -666,6 +666,13 @@ def fetch_models() -> list[str]:
 
 def prepare_responses_body(body: dict[str, Any], *, force_stream: bool) -> dict[str, Any]:
     normalized = dict(body)
+    input_value = normalized.get("input")
+    if isinstance(input_value, str):
+        normalized["input"] = (
+            [{"role": "user", "content": input_value}] if input_value else []
+        )
+    elif not isinstance(input_value, list):
+        normalized["input"] = []
     normalized.setdefault("instructions", "")
     normalized.setdefault("store", False)
     normalized["client_metadata"] = merge_client_metadata(normalized.get("client_metadata"))