Skill UI

This skill provides deep integration with Google NotebookLM, enabling users to query uploaded documentation using Gemini. It ensures source-grounded, citation-backed answers, significantly reducing hallucinations. Features include managing the notebook library (add, list, search), handling authentication, and implementing a structured follow-up mechanism for comprehensive research.

Provides a minimal working example demonstrating how to integrate Perplexity's API for web-grounded answers. Learn to perform advanced searches, retrieve cited sources, and handle features like domain filtering and streaming using both TypeScript and Python. Ideal for integrating real-time, source-backed search functionality into applications.

Leverages firecrawl and exa MCP tools to run focused multi-source searches, scrape key documents, and synthesize citation-backed reports whenever a user needs thorough evidence-based research on any topic.

This meta-skill orchestrates a multi-step process to transform complex research questions, data, or deep reports into a structured, source-backed presentation deck (PPTX). It automatically handles searching, source curation, synthesizing an executive storyline, and generating a professional slide outline and talk track, making it ideal for competitive analysis, leadership briefings, or academic presentations.

This meta-skill is designed to convert complex web research requests into polished, structured, and source-cited deliverables. Use it when you need a market briefing, technical report, or decision memo that requires deep web lookup and scholarly citation. It orchestrates the entire process, including preference inference, multi-source search, drafting, and optional DOCX export, ensuring the output is authoritative and actionable.

This skill performs deep 'git archaeology' by cross-referencing commit history, changelogs, and initial summaries. It extracts verified metrics like commit velocity, contributor rankings, and release timelines, generating multiple detailed reports. Use this when building a REX talk and needing factual, data-backed evidence from a source git repository.

A rigorous framework to prevent confident fabrication (hallucination) by treating competence as an evidence boundary. It guides users to classify claims—Grounded, Partial, or Ungrounded—based on verifiable sources. It mandates assessing the cost of wrongness before answering, fetching, or abstaining, ensuring that assertions are always backed by sufficient evidence.

Provides a rigorous, multi-factor, sector-relative fundamental analysis of publicly listed companies (US, Indian, or global). This skill generates an evidence-backed written report and a sector-relative scorecard, analyzing profitability, returns on capital, balance sheets, and governance. It emphasizes citing primary source documents (e.g., 10-K, annual reports) and benchmarking against appropriate industry peers, avoiding single-metric comparisons.

Review repository architecture using the Ontoly Software Graph and MCP capabilities. Explains module boundaries, package topology, service ownership, dependency cycles, and architectural risk with graph-backed evidence, node ids, edge types, source spans, and confidence levels.

Reviews pull requests using source inspection, static analysis, and evidence-backed AI assistance to report actionable bugs, security risks, architectural issues, and unverified coverage.

Learn to operate GPU-backed Kubernetes clusters for AI inference and training workloads. Covers NVIDIA GPU Operator installation, device plugin configuration, MIG partitioning for resource sharing, and time-slicing for non-MIG GPUs. Includes GPU-aware autoscaling, health monitoring with DCGM and Prometheus, and troubleshooting scheduling or driver issues. Essential for production AI infrastructure management.

Builds a deterministic Software Graph from TypeScript repositories and exposes it via CLI, MCP server, and agent skills for evidence-backed architecture review, request tracing, impact analysis, and dependency topology inspection before searching source files directly.