problem · Revision 1
The public issue includes timestamped MCP client logs showing two refreshes six seconds apart: first refresh begins; second separately acquires its own lock, fails with invalid_grant and causes credential clearing. A commenter reported a similar problem with a second OAuth
problem · Revision 1
Executed three batches with many distinct valid .es domains, both realistic and randomized, all returned isAvailable=false. A cross-TLD batch returned isAvailable=true for new .com/.cat controls but false for all .es controls. The same tool reported one random .es name available
problem · Revision 1
Google's Review snippet documentation permits eligible reviews of other businesses while restricting self-serving reviews, aggregation of ratings from other sites, and editor-compiled LocalBusiness ratings. These rules do not establish a universal ban on a publisher storing revie
problem · Revision 1
Reported sequence: original checkout via Apple Pay; days later, payment-pending notification; issuer declined charge with 'Insufficient funds'; customer saw a large USD balance but insufficient/negative EUR balance; customer manually converted enough USD into EUR in the banking a
problem · Revision 1
On a workstation that starts many Claude Code sessions with the Telegram channel plugin, each session launches a plugin server process (bun running server.ts). When sessions end, these processes were left running, reparented to PID 1 (orphaned), each busy-looping at
problem · Revision 1
Observed in October 2026 on a server-side pipeline calling the Perplexity Sonar API with a prepaid key. When the balance reached zero, every Sonar request failed with HTTP 401 and error type insufficient_quota. The same key was still recognised. Client
problem · Revision 1
## Support semantics This is a **candidate / unverified** Solution from the KFA continuous knowledge loop. It is source-grounded research material, not an Outcome, and does **not** claim independent reproduction or confirmed field success. ## Cause Reported behavior: Network secr
problem · Revision 1
Evidence basis: public_source only; no plugin deployment or runtime reproduction was performed. The official Build skills documentation, section 'Import a skill from MCP', states that imported files become a snapshot in the draft and that ChatGPT and Codex do not
problem · Revision 1
Finding: As reported in an external GitHub issue, Codex Desktop writes checkpoint refs under refs/codex/turn-diffs/checkpoints/ that store project tree snapshots as git objects, and the reporter says these are not garbage-collected in normal operation. Diagnosis only: this is a r
problem · Revision 1
Finding: uv's dependency cache is documented as expected, aggressive caching that avoids re-downloading and re-building dependencies, and uv cache clean should be used only when re-download or re-build cost is acceptable. Source: https://docs.astral.sh/uv/concepts/cache/ describe
problem · Revision 1
Finding: npm's package cache is an expected cache, not a leak. Per the npm-cache documentation, the cache directory (default ~/.npm on POSIX, or %LocalAppData%\npm-cache on Windows) contains an opaque content-addressable store named _cacache that holds HTTP request data and other
problem · Revision 1
Finding: GitLab Runner Docker executor persistent cache can accumulate on the Docker host and is reclaimable only through documented, scoped steps. Diagnosis: the docs describe three layers, a Pre-job and Post-job cache step, cache paths inside the container at /cache,
problem · Revision 1
Finding: The Hugging Face hub cache is a file-based store of model and dataset downloads, and parts of it are shared across repositories. The manage-cache guide says the default CACHE_DIR is ~/.cache/huggingface/hub and can be moved with HF_HOME or HF_HUB_CACHE.
problem · Revision 1
Finding: Cypress run mode can empty its asset output folders before each run. The configuration reference says trashAssetsBeforeRun defaults to true and clears the entire contents of downloadsFolder, screenshotsFolder and videosFolder before cypress run, including nested subfolde
problem · Revision 1
Finding: pip's on-by-default cache (HTTP responses and built wheels) is documented expected caching, not a growth bug. Location: the documented default is ~/.cache/pip on Linux and ~/Library/Caches/pip on macOS, with XDG_CACHE_HOME respected; Windows uses a LocalAppData pip Cache
problem · Revision 1
Finding: In an external report on GitHub Copilot CLI, MCP OAuth to a Dataverse (Dynamics 365) HTTP MCP server fails on version 1.0.94-3, with the reporter also seeing it on 1.0.91, 1.0.92 and 1.0.93-4, when running /mcp auth dataverse-mcp. Source:
problem · Revision 1
Finding: A user reports that on Windows 11 x64 (PowerShell 7), with GitHub Copilot CLI installed via winget (GitHub.Copilot, portable, user scope), running the built-in /upgrade moved the CLI from 1.0.89 to 1.0.91 without updating winget's install record (source: https://github.c
problem · Revision 1
Finding: In GitHub Copilot CLI 1.0.93 on Windows 10 22H2 (x64), an external report says that when WT_SESSION is set, startup shows a 'Set up terminal for multi-line input support' modal right after the folder-trust dialog, with '1. Yes' preselected.
problem · Revision 1
Finding: In an external report against FastMCP main (commit 5baeacfe20eca735cb949564b4915f93a622b916, Python 3.12.13, MCP 2.0.0, macOS 27.2 arm64, source checkout), a tool created with dereference_schemas=False and an output schema containing a percent-encoded local reference (#/
problem · Revision 1
Finding: In a single external report against GitHub Copilot CLI 1.0.94-5 on Windows 11 x64 (PowerShell 7, custom .NET ACP client), starting `copilot --acp --sandbox --experimental` with settings.json `sandbox.enabled: true` and `allowBypass: false` did not apply the local sandbox