⚬ PROBLEM: A recent change in YouTube's desktop UI, removing the sub scount found below the creator's name when watching a video. ⚬ SOLUTION: <!-- (Describe what you'd like (A clear and concise description of what y

https://meshery.io/extensions is missing an entry for https://github.com/tata-consulting/meshery-remote-provider The meshery.io website uses Jekyll and GitHub Pages. Site content is found under the master branch. - 📚 See the [Contributing to Meshery.io Website](https://github.com/meshery/meshery.io#contributing-to-the-mesh

On the Meshery website, under Getting Started, the “Meshery Docs” link is not highlighted, so it is not visually prominent and does not clearly look like a clickable link. It should be highlighted like the “Extension Points” text in the other section, so users can easily identify it as a hyperlink. The “Meshery Docs” text should be clearly highlighted and visible as a clickable link, similar to “Extension Points”. |

macOS v1.4.146 When searching for files by name in a git repo directory, the search fails with the error "File listing exceeded 10000 files". This occurs even when searching for a specific, known file name (single file). I found that the file listing strategy depends on whether ripgrep is installed. Initially, the search failed with the error "File listing exceeded 10000 files" even when searching for a specific

The goal of this issue is to improve the Spark documentation by restructuring it into multiple sections, similar to the Trainer and Optimizer documentation. Potential improvements include: - Add a Example section. - Split the documentation into dedicated guides instead of a single page. - Improve the current documentation This issue can be implemented through multiple documentation PRs. The Spark documentation currently consists of

Open Design's stdio MCP server advertises capabilities.resources: {}, but it does not implement the resources/templates/list method. MCP clients that enumerate the complete advertised resource surface receive JSON-RPC error -32601 Method not found. I reproduced this with the packaged macOS app, and the handler also appears to be missing from the current main branch. - Open Design desktop: 0.16.1 - OS: macOS - Transport: stdio - MCP client SDK:

What do we need to build or fix? We need to update footer to include link to the new service page, there may also be some component styling update - please check figma Technical details Footer on figma https://www.figma.com/design/Fp0lLB5OWZLJseTs2svGbg/Web-Dev-Path_design?node-id=85-2&p=f&t=Tt1geRvziQZqikkV-0 Approach suggestions 1. Check our service page google docs for detailed workflow, make sure you branch from the latest services branch services-dev 2. PR should be made to

Omnigent can express agents, tools, policies, sub-agents, and handoffs. A multi-stage workflow is still encoded mainly in an agent prompt and executed through model-chosen tool calls. Prerequisites, required outputs, pause points, agent changes, and completion conditions are not declared as one validated workflow. This makes a repeatable workflow hard to build and review. Our developer narrative requires these stages: 1. Prove that AI Gateway blocks unsafe input and per

FluxCast now supports easy localization, and we want to expand beyond English, Russian, and Czech! If you want to see the system tray interface in your native language, you don't need to write any Python code. 1. Open src/i18n/translations.json. 2. Add your language code (e.g., "de", "es", "fr") and the translated text to the existing keys. 3. Open a Pull Request! It takes less than 5 minutes as there are only ~30 strings to translate. Than

PR #460 stood up the new docs site (pydata theme, d2 diagrams, tested examples) as a scaffold + middle ground — the structure, wiring, and current-API correctness are in, but substantial prose still needs filling / refining across the guides. This issue tracks that content pass so #460 can land immediately as the scaffold. Refine + flesh out the written content across the docs tree (docs/). Fill in per-page bullet notes as the refinement passes happen: - [

Item 6 from #143, deferred as a design decision. SKILL.md's self-reference escape hatch says text inside quotation marks, code blocks, or explicitly marked as illustrative is exempt from flagging. scripts/self-scan.js implements exactly that with a QUOTED_SPAN regex. The engine does not: it masks only multi-line blockquotes (two or more consecutive > lines). ` $ node -e " const D=require('./detector/patterns.js'); const t='She told me, \"We must delve into the landscape and leverage o

1. Start Build a Prime Number Checker Module. 2. Create index.js with: 3. Add a valid CommonJS package.json. 4. Click Run Tests. The page reports: ` Error 4XX - 5XX Cli

I think that it would be great it we could get some more output on the console when AutoIt crashes. For example, adding the 0x hex code and NTSTATUS description. This would be much more beneficial when troubleshooting issues with a script. For example, we currently get something like: AutoIt3 ended. rc:-1073740771 But instead, it would be awesome if we could get: `AutoIt3 ended. rc:-10737407

Seeing a lot of traffic by nbot suspect it is an AI crawler that might need adding to the list

We already have a good tutorial for Cartesian Histograms at https://www.pygmt.org/0.19.0/tutorials/advanced/cartesian_histograms.html, which was initially developed more than 3 years ago (#2445). Here are some ideas to improve the tutorial: - The series parameter is one of the most important parameters that controls the histogram, but it's not explained well. The first subsection should focus on this parameter - Inspired by [matplotlib's hist method's histype parameter](https://matplotli

There is no way to tell any of these engines "I know about this one, stop telling me". That is the thing that stops a scanner being adopted on an existing codebase. Point secret-scanner at a repository with a test fixture that looks like a key, or sast-lite at code with a deliberate eval behind a comment, and the run fails. The only lever today is --min-severity, which is a blunt instrument: raising it to hide one known finding hides every real finding at that level too. What to do: -

Have this Mac? You can contribute useful benchmark data without changing any code. Every number laya-apple publishes comes from one Apple M4 Max. An M3 Max result shows how the MLX backend performs on the previous generation, and whether the Neural Engine parity gate passes there. | | | |---|---| | Hardware needed | An M3 Max Mac (any memory size) | | Estimated time | About 10 minutes for an MLX-only result, about 25 minutes with the optional ANE step | | Code changes | Non

The job-started and job-completed hooks can hold a dispatched build. Their two existing suites cover slot accounting, limits, stale-slot reaping, fail-open behavior, and installation against a fake fleet, but the shell-tests CI job does not run them today. Wire the existing suites into the shell-tests job in .github/workflows/ci.yml. This is CI wiring only; do not change hook behavior. Suites: - hooks/tests/run.sh - hooks/tests/install.sh

Add header/main/footer, alt text, labels, focus styles. AC: Lighthouse a11y 90+.

The README still lists “IMPORTANT: WRITE TESTS” as unfinished. The current lint workflow runs Ruff checks and formatting only, so it cannot detect incorrect analysis results such as state leakage or wrong metric aggregation. Sources: README, lint workflow. This issue is based on repository inspection; no test suite has been executed as part of this re

The /components folder will contain all the generic components that will be used in various parts of the application. The components will be (for now): - ToolCard - ToolForm - ToolFilters - StarButton - LoginForm - RegisterForm - AppHeader - AppFooter

Measure and document the exact gas consumption and transaction cost () of Werracle.sol forward decision calls (decideNoul, decideChoice) across major Ethereum Layer-2 rollups.

From the survey in #274, of 212 co-located rows in 184 groups: | source | duplicate rows | share of that sou

Let users choose how long completed-task history is kept instead of retaining every entry indefinitely. The monitor is local-first and may stay running for months. A small retention policy keeps the local data predictable without requiring manual cleanup. - Add an environment variable such as COMPLETIONRETENTIONDAYS. - Keep the current behavior as the documented default, or choose a conservative default and explain it. - Remove expired

The VS Code extension opens the tour and jumps from a component to code. The reverse direction is missing, and it is not published. Add a command that finds the component containing the active file (using each component's sources) and reveals and selects it in the tour panel. Then prepare publication: an icon, a Marketplace description, vsce packaging in CI, and a publish workflow gated on a

pluk version prints a hardcoded string in src/cli.ts: package.json is at version 0.8.6, so the CLI misreports its version by seven releases. This also affects pluk attach's dependency probing UX (resolvePlukBin runs ... version) and any operator triage that trusts the reported version. Read the version from the package's own package.json at runtime (it is adjacent to dist/ and

Gate 1 refuses to call the answer model when no retrieved passage clears a cosine-similarity threshold, tau_min = 0.35. This value is not calibrated. config/retrieval.yaml states so explicitly: The same applies to

Internationalization / translations Every word of marketing copy currently lives in frontend/messages/*.json — 13 files, ~700 lines each. Changing a headline means editing JSON in 13 languages, opening a PR, passing CI and waiting for a deploy. Nobody outside the engineering team can change a sentence on the homepage, and the rebrand is about to rewrite a lot of that copy. The translation side is the harder half. The site supports 13 locales: `en, fr, sw

- Play-testing show health too low - Marisa one shots everyone - Add more varieties to skill, right now just all do 4 damage for tier 2 - Tier 3 skills? - Spell card variety

Agents and web servers built on asyncio (FastAPI and similar) shouldn't block the event loop on the routing call. Add await router.aroute(task, limits=None), starting with asyncio.to_thread(self.route, ...). This keeps zero dependencies and fits the small codebase. A native async HTTP client can come later if someone needs it.