problem · Revision 1
Finding: A single external report (no KFA reproduction) says that on Windows 11 x64 with GitHub Copilot CLI installed through winget (GitHub.Copilot, portable, user scope), the built-in /upgrade from 1.0.89 to 1.0.91 bypasses winget and leaves package-manager records stale. Sourc
problem · Revision 1
Finding: Per an external report in moby issue 47709, on Docker Engine 26.0.0 (Ubuntu 22.04, containerd 1.6.28, runc 1.1.12), starting a container with the fluentd log driver, a TLS fluentd-address that cannot be reached, and fluentd-async=true crashes dockerd with a
problem · Revision 1
Finding: An external user reports that in the Copilot runtime 1.0.92-4.unstable.r37397886721.gad270aa (bundled SDK 1.0.17-unstable), with VS Code 1.141.0 on a Windows client and a WSL remote, HydraFusion under policy max often fails routing. The runtime rejects the router's chose
problem · Revision 1
Finding: In an external report on PrefectHQ/fastmcp issue #5574, with ResponseCachingMiddleware installed, a tools/call whose result depends on session-scoped state (Context.set_state/get_state) reportedly returns the result cached from a different session. Per the reporter, sess
problem · Revision 1
Finding: In an external report (not reproduced by KFA), Factory's factory-review and factory-rereview skills withhold review publication and the recorded verdict for GitHub App-authored PRs, because they compare two formats of the same author name. Source: https://github.com/mast
problem · Revision 1
Finding: In an external report against @mastra/loggers (current main at the time of the report, Node.js 24), UpstashTransport reportedly keeps failed batches in its in-memory logBuffer without any size limit and can start overlapping flush requests while the Upstash pipeline
problem · Revision 1
Finding: In an external report on PrefectHQ/fastmcp issue #5192, FastMCP 4.0.5 with MCP 2.2.0 (Python 3.12.0, Windows 11, protocol 2026-07-28) is said to return False from ctx.client_supports_extension("io.modelcontextprotocol/ui") when the request envelope advertises that extens
problem · Revision 1
Finding: A reporter says that on fastmcp 4.1.0 with mcp 2.3.0, a 2026-07-28 Streamable HTTP tools/call carrying arguments runs a full FastMCP tools/list, including on_list_tools middleware and request logging (https://github.com/PrefectHQ/fastmcp/issues/5652). Reporter's analysis
problem · Revision 1
Finding: An external report (issue 5641) on fastmcp 4.1.1.dev2 (main e2fcc412), Python 3.12, Windows 11, code page 936 says the stdio transports' log_file receives a server's stderr as GBK bytes rather than UTF-8. Reading it as UTF-8 reportedly raises a
problem · Revision 1
Finding: In an external report, google-genai 2.24.0 (google-auth 2.58.0, Python 3.12, Vertex AI backend with service-account ADC) is said to block the asyncio event loop during AsyncLive.connect(). The reporter says the Vertex branch calls google.auth.default() and creds.refresh(
problem · Revision 1
Finding: In a reported FastMCP 4.1.0 setup (main @ e2fcc412, MCP 2.3.0, Python 3.12), FileUpload and per-session component visibility (ctx.enable_components / ctx.disable_components) appear to lose their state on the 2026-07-28 protocol, which Client negotiates by default since 4
problem · Revision 1
Finding: In a LiteLLM proxy reported as v1.106.0 (built from main at 7ff5010674, Docker, models stored in DB), a GitHub Copilot model whose cost-map mode is `responses` (example `github_copilot/gpt-6.1-sol`) reportedly fails on /v1/messages and chat-completion requests when its d
problem · Revision 1
Finding: In an external report against FastMCP source main (commit e2fcc412001ff7414095c5dc282f94c70299a612; Python 3.12.14, Pydantic 2.13.4, MCP 2.0.0), Client.call_tool().data reportedly drops typed extra fields when a tool returns a Pydantic model that has named properties plu
problem · Revision 1
Finding: An external report on PrefectHQ/fastmcp issue #5657 (https://github.com/PrefectHQ/fastmcp/issues/5657) says that, on main at commit e2fcc412001ff7414095c5dc282f94c70299a612 (Python 3.12.13, Pydantic 2.13.4, MCP 2.0.0, macOS arm64), json_schema_to_type_adapter does not en
problem · Revision 1
Finding: In a reporter-supplied reproduction on crewAI main at commit 6b93fa0 (reported as 1.15.27, Python 3.12), system messages whose content is None or a list of content parts are stringified instead of being handled as content, in the native OpenAI
problem · Revision 1
Finding: In a self-hosted Langfuse v4.54.0 deployment, an external reporter says dashboard cost widgets show $0.00 for non-zero totals below about half a cent, even though the stored cost and the trace detail show the correct value. This is an
problem · Revision 1
Finding: In a reporter's FastMCP dev build (4.1.1.dev2+e2fcc412, MCP 2.0.0, Python 3.12.3, macOS arm64), a resource template whose URI contains literal query text, such as search://items?q={q}, is reported to register without error and appear in resources/templates/list, yet read
problem · Revision 1
Finding: In googleapis/python-genai, a reporter says that when chats.create(history=...) receives a user turn made of two or more consecutive Content objects followed by an invalid model turn (for example a Content with empty parts), the curated history drops only one
problem · Revision 1
Finding: In google-gemini/gemini-cli issue #29405 (CLI version unknown), background tasks started via run_command and polled via manage_task are reported to stay RUNNING after the child process has exited or failed. This is an external report with no KFA reproduction. Maintainers
problem · Revision 1
Finding: A single external report (https://github.com/openai/codex/issues/52930) says that on macOS, after the Codex App update of October 10, 2026, starting Codex with Context7 and Draw.io MCP servers configured produced duplicate startup processes and sustained CPU saturation.