Skill UI

Provides calibration-free weight quantization for LLMs down to 1-bit precision with multiple optimized backends, ideal for fast deployments with vLLM/HuggingFace and LoRA/PEFT fine-tuning without collecting calibration data.

This skill provides a comprehensive framework for performing end-to-end UI testing and verification on Android emulators. It guides users through systematic processes including device calibration, UI element discovery using uiautomator, performing complex interactions (tap, swipe, input), and robust reporting via visual screenshots and logcat analysis. Ideal for ensuring the stability and functional correctness of React Native or native Android applications.

This comprehensive guide details the use of FFUF, a fast web fuzzer written in Go. It is designed for advanced security penetration testing, providing techniques for content discovery, subdomain enumeration, parameter fuzzing, and POST data testing. Learn to master the FUZZ keyword, advanced filtering, and auto-calibration to efficiently identify hidden paths and potential vulnerabilities in web applications.

This methodology provides a structured framework for stress-testing critical business assumptions before they fail in the market. Learn to systematically identify, challenge, and model the downside risk associated with market size, revenue projections, competitive moats, and customer behavior. It moves beyond optimism to achieve crucial business calibration.

This skill generates highly structured DESIGN.md files, acting as a semantic design system for Google Stitch. It translates complex, anti-generic frontend principles—including typography, color calibration, component behavior, and motion physics—into actionable, natural-language rules. Use it as the single source of truth to guide AI in generating premium, non-clichéd, professional user interfaces.

Generates structured DESIGN.md files that serve as a single source of truth for AI-powered screen generation (Google Stitch). This skill translates complex, anti-generic design directives into precise semantic rules, covering typography, color calibration, motion physics, and layout principles to enforce a premium, high-agency UI/UX standard. It ensures outputs are sophisticated and avoid common AI design clichés.

A powerful 12-agent pipeline designed for academic writers, guiding users through the entire research publication lifecycle. It covers systematic literature review, structural outlining, full drafting, citation compliance, bilingual abstract generation, and simulating peer review. Features include Style Calibration and Writing Quality Checks to ensure high academic rigor and professional polish.

This agent simulates a comprehensive international journal peer review process. It dynamically deploys five expert reviewers—including a Journal-Fit Analyst and a Devil's Advocate—to assess a paper from multiple angles (methodology, domain, originality, logical fallacies). It generates structured reports, an Editorial Decision, and a detailed Revision Roadmap, supporting full reviews, re-reviews, and calibration checks.

This framework serves as a rigorous thought partner, designed to challenge underlying assumptions, separate raw evidence from interpretation, and ensure that all conclusions are fully defensible. It guides users to define clear decision goals, test counterfactuals, and calibrate claims strictly to the available data quality. Ideal for high-stakes projects where the cost of error is significant.

This tool analyzes historical sprint velocity data to generate a comprehensive engineering team health report. It moves beyond simple charts by diagnosing underlying delivery patterns, identifying risks like chronic over-commitment, and assessing story point calibration. The report provides actionable recommendations for improving capacity planning and team predictability.

This skill provides utilities to manage and drain the deliberation queue (brain_frames) of the ACC system. It allows users to list waiting frames, retrieve necessary context, perform deliberation on open tasks, and submit continuations using `acc_act(runtime="continue")`. It is essential for checkpointed work saved by headless runs, ensuring continuous task resolution and calibration against predicted outcomes in complex AI workflows.

This example demonstrates unsafe or malicious review data inputs, specifically designed for calibration and security testing. It simulates prompt injection attacks and must only be used in controlled environments to validate secure data processing and system robustness.