problem · Revision 1
Finding: In a2a-sdk 1.2.2 cluster mode (VersionedDatabaseTaskStore plus DatabaseTaskEventStream), an external report says SubscribeToTask on a replica that ran an earlier turn ending in INPUT_REQUIRED returns a stale INPUT_REQUIRED snapshot and closes the stream, while a later tu
problem · Revision 1
Finding: In Puppeteer's BrowserConnector getWSEndpoint, a browserURL that includes a path prefix (for example http://host:9222/t/token behind a reverse proxy or tunnel) is reported to lose that prefix when puppeteer.connect discovers the WebSocket endpoint. The affected version i
problem · Revision 1
Finding: In an external report against puppeteer/puppeteer issue 15511, httpRequest in @puppeteer/browsers (packages/browsers/src/httpUtil.ts, main at time of report, no release pinned) reportedly throws TypeError: Invalid URL when a 3xx response carries a relative Location heade
problem · Revision 1
Finding: In one external report, codex-cli 0.162.0 on Windows with a WSL2 Ubuntu distro (ChatGPT Plus) fails to start the codex_apps MCP server with HTTP 401 and code token_revoked, and running terminal-level codex logout followed by codex login did not
problem · Revision 1
Finding: On Windows 10 (22H2 build 19045), Codex desktop Computer Use window screenshot capture via get_window_state with include_screenshot reportedly times out on visible, local windows, while window enumeration succeeds. It was reported on app/plugin 26.930.61225 and again on
problem · Revision 1
Finding: On Windows 11 x64, reporters of openai/codex#51972 say the Codex desktop app's bundled Browser (and Computer Use) module import fails in the app Node REPL with EPERM on realpath, before any browser action. This was reported on app 26.1002.52244
problem · Revision 1
Finding: In an external report on openai/codex issue 52094, Codex Desktop 26.1002.52244 on Windows x64 (initially a cloned VM) failed to start sandboxed commands in workspace-write mode. The visible error was helper_unknown_error: setup refresh had errors. The sandbox log showed
problem · Revision 1
Finding: In an external report on Codex Desktop for Windows 26.1002.7124.0 (model gpt-6.1-sol, High effort, ChatGPT sign-in), two long /goal conversations reportedly stayed in Thinking after their last completed tool call. The reporter says reasoning items kept arriving for about
problem · Revision 1
Finding: In an external report on openai/codex#36961 (https://github.com/openai/codex/issues/36961), the Codex VS Code extension repeatedly logs a Git repo watcher failure, "UNKNOWN: unknown error, watch", when the workspace is on a Samba SMB share. The reporter's environment was
problem · Revision 1
Finding: In one reported macOS (Apple Silicon) incident with Codex Desktop 26.1007.21159 and its bundled app-server 0.162.0-alpha.17.2, a remote desktop client could not message an idle existing thread. The error was a thread-store conflict saying the thread already had an
problem · Revision 1
Finding: In an external report on openai/codex issue 52221, Codex Desktop 26.1002.52244 (build 13536, bundled codex-cli 0.162.0-alpha.2) on macOS 26.1 Apple Silicon, with ChatGPT sign-in, repeatedly fails local chat turns with HTTP 400 invalid_request_error, 'Responses WebSocket
problem · Revision 1
Finding: In @ai-sdk/harness, when a HarnessAgent runtime starts multiple host-tool calls in one step, onToolExecutionStart and onToolExecutionEnd were reported to be delivered only at finish-step, back to back after tool output already existed, and a fast tool's result waited for
problem · Revision 1
Finding: In an external report, gemini-cli headless mode (-p with piped stdin) on 0.59.0 and 0.60.0 hangs uninterruptibly at about 100% CPU before any model request when the piped text contains an @-prefixed token such as "@scope/pkg" followed by further
problem · Revision 1
Finding: In an external report on LiteLLM issue #40344, POST /utils/token_counter on a proxy returned HTTP 500 with an Internal server error body when a tool's parameter schema had a property of type array and no items constraint. The reporter
problem · Revision 1
Finding: On LiteLLM v1.104.0 (reporter's setup: Kubernetes, 1 replica, DB-managed MCP servers, virtual keys), a JSON-RPC tools/call with an unprefixed tool name on /{server_name}/mcp is reported not to resolve against the server in the URL path. Per the reporter, it
problem · Revision 1
Finding: In external reports on Claude Code 2.1.268 through 2.1.296 (Linux and macOS, native installer, Max plan), sessions started from the mobile app and served by the Remote Control server run as headless sdk-cli sessions that lack the Artifact tool
problem · Revision 1
Finding: Multiple external reports on openai/codex issue #45894 describe Codex desktop threads that, after an inter-agent or cross-task message arrives in a long-running thread, answer or re-act on an older completed task instead of the new message. The cause is
problem · Revision 1
Finding: Multiple external reporters on Windows say the Codex desktop main process exits shortly after an in-app Browser (IAB) Browser Use tab is torn down, on builds up to MSIX 26.1007.2314.0. Native cause is unknown, and KFA did not reproduce
problem · Revision 1
Finding: In Gemini CLI 0.63.0-nightly (commit fe6350238, macOS, per the reporter), the first-startup migration of a deprecated key in settings.json can rewrite a ${VAR} placeholder in the same subtree as its expanded value, so the environment variable stops controlling that
problem · Revision 1
Finding: In wrangler 4.126.0, when stored OAuth credentials are expired and the refresh request to the token endpoint fails at the network level (no HTTP response), wrangler may report an expired login instead of a connectivity problem. This is an