metabench
Community@metabench
metabench provides skills for building, verifying, and styling jsgui3 SSR UI components, SVG design systems, and disciplined agent research loops.
Agent Skills by metabench
Showing 38 vetted skills indexed across 3 GitHub repositories.
metabenchjsgui3-verify
Verifies jsgui3 SSR and client reattachment behavior through runtime measurement instead of source inference.
metabenchwlilo
Generate luxury-styled SVG diagrams, posters, and CSS themes using the WLILO design language.
metabenchsingularity
Runs a bounded recursive improvement cycle over agent instructions and the news crawler codebase.
metabenchloop-audit
Audit self-improvement loop cycles against git history and ledger claims.
metabenchjsgui3-context-menu-patterns
Implements context menus in jsgui3 applications with positioning, clamping, and dismissal handling.
metabenchui-screenshot-feedback
Automates UI screenshot capture and comment-driven iteration using Puppeteer, Electron, or Playwright.
metabenchexperimental-research-metacognition
Guides hypothesis-driven experiments with minimal reproductions to validate uncertain subsystem behavior.
metabenchwlilo-design-system
Applies WLILO design tokens across HTML, SVG, and jsgui3 interfaces.
metabenchjsgui3-lab-experimentation
Validates jsgui3 behavior questions through lab experiments and distills findings into documentation.
metabenchsvg-collisions
Detect and repair overlapping text and shapes in SVG diagrams.
metabenchsvg-theme-system
Generates themeable SVG diagrams using shared design tokens and selectable palettes.
metabenchjsgui3-wlilo-ui
Applies WLILO design tokens and styling conventions to jsgui3 UI components.
metabenchjsgui3-ssr-activation-data-bridge
Diagnose jsgui3 server-rendered markup hydration and client-side control activation issues.
metabenchmcp-memory-server-surgery
Guides safe modification of MCP servers exposing repo memory with protocol guardrails and validation steps.
metabenchtelemetry-contracts
Maintain telemetry contract stability across event names, JSONL envelopes, and status endpoints.
metabenchpuppeteer-efficient-ui-verification
Verify UI scenarios with Puppeteer using reusable session-based checks.
metabenchtargeted-testing
Select and execute the most concise validation step for code modifications.
metabenchjsgui3-context-menu-patterns
Implement context menus in jsgui3 applications with viewport clamping and dismissal.
metabenchunderstanding-jsgui3
Guide building and debugging jsgui3 UI components with server-side rendering.
metabenchexploring-other-codebases
Map unfamiliar codebases by identifying foundational files and progressively disclosing content.
metabenchagent-output-control
Manage agent output verbosity, terminal modes, and emoji-based logging.
metabenchdeep-research
Conduct iterative multi-stage research with evidence grading and reasoning loops.
metabenchrunaway-process-guard
Enforce timeouts and termination protocols to prevent AI agent process deadlocks.
metabenchskill-writing
Guide creation and auditing of Antigravity Skills with structured procedures.
Frequently Asked Questions About metabench
FAQPage SchemaWhat tasks can I accomplish with metabench's skills?▼
You can build and debug jsgui3 UI controls with server-side rendering, verify UI changes via Puppeteer screenshots and bounding-box metrics, create WLILO-styled SVGs and CSS themes, repair SVG collisions, run targeted tests and static analysis, and manage long-running agent sessions with ledgers and state files.
Who are metabench's skills designed for?▼
They target developers and AI agents working in the jsgui3 ecosystem—engineers building SSR controls, debugging client activation, styling dashboards in the WLILO design language, or operating recursive self-improvement loops like singularity and loop-audit on projects such as copilot-dl-news.
How do I verify a jsgui3 UI change actually works?▼
Use jsgui3-verify and browser-verification: launch demo servers, render in a live browser, capture screenshots, and check client-side activation via the SSR data bridge. The skills warn against trusting green test suites when the page behavior disagrees, and gate 'verified' claims in commit messages.
What prerequisites do these skills assume?▼
Skills assume a JavaScript/TypeScript environment with jsgui3 components under src/ui, Puppeteer or Electron for capture rigging, and repo structures like src/ui/lab for experiments. MCP memory server surgery additionally requires familiarity with stdio JSON-RPC, tools/list, and tools/call protocol details.
How do the singularity and loop-audit skills work together?▼
Singularity runs one bounded, verification-gated cycle that improves agent instructions and implements crawler plus jsgui3 UI work. Loop-audit then judges those cycles—checking convergence versus busy-work, substantiating ledger claims against git, and validating tech-tree accuracy before deciding whether the loop should continue.