Skill UI

A comprehensive suite designed for AI agents to perform safe, deep, and multi-format web searches. It allows agents to retrieve current web information, analyze PDFs, extract YouTube transcripts, and access GitHub code files. Instructions include guidelines for varying search depths and using specialized tools for code context retrieval.

This skill extracts structured data, such as text, tables, or attribute values, from rendered web pages. It supports both running pre-saved templates and conducting one-shot DOM queries. Crucially, it integrates mandatory AIDefence gates for PII scanning and prompt-injection detection, ensuring data safety and reliability before the data reaches the model.

This tool performs robust outlier detection on individual session spending. Utilizing Median and Median Absolute Deviation (MAD), it calculates a modified z-score, making it highly resistant to extreme values, unlike mean+sigma methods. It specifically identifies which singular sessions are anomalous, making it crucial for cost monitoring, auditing, and integrating into CI/CD pipelines to prevent unexpected resource overruns.

This skill provides a granular, session-by-session view of total accumulated costs. It lists metrics such as start time, message count, top model, and total cost for every recorded conversation. This is ideal for identifying which specific sessions consumed the most resources, which is crucial for project retrospectives and evaluating the sustainability of current session patterns.

This tool provides forward-looking cost projection by calculating the average daily burn rate from a specified measurement window. It linearly extrapolates future spending across multiple time horizons (7d, 30d, 90d, 365d) and alerts users to the projected date of budget exhaustion. It is essential for proactive financial planning, SRE cost monitoring, and setting CI/CD build cost gates.

This tool provides a granular, per-message cost breakdown within a single conversation session. It serves as a deep-dive utility following an anomaly detection, allowing operators to pinpoint exactly which messages incurred the highest costs. It reveals critical cost components like output tokens, cache writes, and model escalations, helping users understand the true resource consumption beyond simple token counts.

Build a production-ready, AI-powered data collection agent for any public source (job boards, prices, news, etc.). This agent automatically scrapes data, enriches it using free LLMs (like Gemini Flash), and stores the results in databases such as Notion, Sheets, or Supabase. It runs on a free schedule using GitHub Actions and learns from user feedback, making it ideal for continuous data monitoring and tracking.

A comprehensive methodology for accelerating large-scale data ingestion, ETL processes, backfilling, and data warehouse loading. It guides users in identifying and optimizing bottlenecks—such as source extraction, network transfer, and loading speed—while strictly ensuring data correctness and maintaining clear accounting checkpoints throughout the pipeline.

A read-only tool designed to compare prediction-market baskets or themes against the user's existing knowledge base, portfolio notes, research memos, or financial context. It performs gap analysis, flagging potential conflicts, stale assumptions, and areas of missing research. This skill provides structured insights into fit and exposure without offering any investment or trading advice.

This skill provides comprehensive, read-only market intelligence for prediction markets. It helps users research market events, compare different venues (like Polymarket, Kalshi), analyze liquidity, and understand underlying assets. It generates source-grounded briefings, emphasizing context and data quality, and strictly avoids providing investment or trading advice.

A comprehensive skill designed to fetch raw market data (Open, High, Low, Close, Volume) for a given symbol. It normalizes the time series data into relative OHLCV vectors, converts these vectors into 64-dimension embeddings, and stores them in a memory store. Crucially, it indexes the resulting vectors using the HNSW algorithm, enabling fast and precise nearest-neighbor search for pattern detection and similarity analysis in financial time series.

This powerful tool scans ingested OHLCV market data to detect and classify known candlestick patterns, such as Doji, Hammer, Engulfing, and Head & Shoulders. It analyzes single, multi-, and complex candle sequences, assigning a pattern name, type (reversal/continuation), direction, and a reliability score (0.0-1.0). Ideal for technical traders needing structured pattern identification for advanced market analysis.