Data analysis, machine learning, LLM apps, and prompt design (9.1K skills)
Guide model selection around service availability, starting with local models and switching only after infrastructure failures. Includes provider health checks, spending checks, and strict limits on when paid models may be used.
Manage the information an AI coding session carries by saving notes, selecting relevant material, summarizing results, and separating tasks. The skill guides context budgets and session handoffs so important decisions remain available during long projects.
Reduce unnecessary output and repeated work with concise response rules, tool-call budgets, and read-before-edit checks. Choose formats for coding, automated pipelines, or analysis, and limit repeated attempts at fixing the same failure.
Refine a SKILL.md using accumulated corrections. Language models propose and evaluate edits against reserved validation examples, accepting only candidates that meet the score threshold while enforcing spending and length limits. The process changes skill instructions without retraining a model.
Review plans and design decisions with several AI models. Each answers independently, ranks anonymous peer responses, and passes the results to a chair model for synthesis, with the full discussion saved for later reference.
Measure an Attio integration’s requests, retries, query patterns, sync churn, and retained data to find avoidable workload. Get ranked, reversible improvement proposals with estimated impact and tradeoffs for data freshness, reliability, and complexity.
Use this skill to build or review live transcription sessions with AssemblyAI Streaming v3. It covers safe authentication, correctly paced audio, partial and finalized turns, reconnect boundaries, and explicit session termination so integrations can handle results and session costs clearly.
Use this skill to measure and tune an AssemblyAI transcription workflow for latency or throughput goals. It helps compare models, audio delivery, polling, streaming turns, concurrency, and downstream work while keeping accuracy, privacy, and cost in view.
Use this skill to add application-level rules around Claude API requests and responses. It covers input checks, sensitive-output filtering, topic and model limits, per-user and global budgets, and safe policy testing with synthetic cases and aggregate records.
Use the Apify JavaScript client to manage datasets, key-value stores, and request queues. Export scraped records as CSV, JSON, or XLSX, keep files and configuration, resume queued requests, and chain Actors into scrape, transform, and export pipelines while monitoring run status and cost.
Add observability to Claude API integrations by recording request metadata, latency, token totals, estimated cost, errors, and rate-limit headroom. It helps define dashboards and alerts, then verify them with synthetic requests while keeping prompts and responses out of telemetry.
Measure and improve Claude API performance through prompt caching, model selection, streaming, and controlled parallel requests. It compares latency, token use, throughput, and quality to document which configuration changes help.
Analyze Claude API spending and test cost controls through model selection, caching repeated input, batch processing, and output limits. Track budgets and compare cost, quality, and response time before expanding a proposed change.
Analyze liquidity pools on Uniswap, Curve, and Balancer using deposits, trading volume, and fees. Estimate returns from supplying funds, compare price-change losses against holding tokens, and export pool comparisons as JSON or CSV.
Analyze text with natural language processing, a set of methods for finding patterns in language. The skill can identify sentiment, extract keywords and named entities such as people or organizations, and surface the main topics in reviews, articles, or other text.
Explore DeFi and blockchain activity with data on protocol and chain TVL, fees, DEX volumes, yields, and stablecoin market caps. Filter results by chain or category, track TVL changes, and export reports as JSON or CSV for further analysis.
Collect a reviewed diagnostic bundle for persistent Claude API problems or a support ticket. It records request IDs, rate-limit headers, SDK and runtime versions, and connectivity details while excluding API keys and customer prompt or response content.
Assess crypto market mood with the Fear & Greed Index, news keywords, and price and volume momentum. Get a 0–100 composite score and component breakdown for the overall market or a selected coin, with options to view trends or export JSON.
Analyze NFT collections by measuring how uncommon each trait is. Calculate rarity scores, rank tokens, compare individual NFTs, inspect trait distributions, and export rankings as JSON or CSV for further review.
Analyze customer reviews, social media posts, and survey responses to classify their tone as positive, negative, or neutral. Sentiment scores help organize feedback and understand the opinions expressed in the supplied text.
Inspect pending blockchain transactions, compare gas fees, and detect swaps on decentralized exchanges. Get transaction summaries and fee recommendations, or study potential MEV opportunities, where transaction ordering can create profit.
Set up privacy practices for Claude API workflows, including redacting personal data before requests, scanning responses, and logging request metadata without message content. Produce a sanitized handling record with retention and deletion outcomes.
Analyze BTC and ETH options activity from crypto derivatives exchanges such as Deribit, OKX, and Bybit. Review open interest, put/call ratios, large trades, implied volatility, and unusual positioning, then export findings as JSON or CSV.
Investigate difficult Claude API failures by isolating token limits, tool schemas, streaming behavior, stop reasons, or prompt settings. Compare a small, safe reproduction with a known-good baseline and produce redacted evidence for review.