Data analysis, machine learning, LLM apps, and prompt design (8.9K skills)
Develop machine learning models from local data, covering data preparation, model selection, training, evaluation, and deployment preparation. Use it to build workflows for tasks such as predicting customer churn or classifying images.
Coordinate multiple AI agents in a mesh, a network where peers communicate directly. It guides shared decisions, task assignment based on skills and workload, network monitoring, and recovery when agents fail or connections break.
Preserve project decisions and work progress across sessions, and share information between agents. Organize stored knowledge into separate spaces, retrieve relevant records, and manage expiration, compression, and cleanup of saved data.
Find stored Ruflo memories by meaning, keywords, or an optional ranking method. Search across storage groups, retrieve past solutions and task context, and see which search method actually ran, including any fallback.
Compare ruflo and HAL systems for running the GAIA benchmark, an AI evaluation. Examine differences in code execution, browsing, file handling, memory, and model selection to prioritize improvements by expected impact and effort.
Guide GAIA benchmark runs from environment checks and cost estimates to result comparison. Create a signed submission package compatible with HAL, including agent traces, metadata, and an audit report that checks for benchmark exploitation.
Guide market prediction workflows with LSTM, Transformer, and N-BEATS models using neural-trader commands and result storage. It explicitly identifies version 2.8.11 outputs as placeholders, so they cannot be treated as fitted models or evidence of prediction accuracy.
Import Claude Code auto-memory into AgentDB and search past notes by meaning across sessions and projects. The skill checks import status, consolidates near-duplicate entries, and returns search results with their memory sources identified.
Represent dependency trees, taxonomies, and other hierarchies inside a Poincare ball, a geometric model for hierarchical distances. Use tree depth for radius and text embeddings for direction, then measure distances to examine the encoded structure.
Store useful patterns and solutions in AgentDB, then find them through semantic search, which matches meaning rather than exact wording. Use saved knowledge before starting work and record successful approaches and new lessons afterward.
Test trading strategies against historical market data and collect performance metrics. The skill checks whether data is real or synthetic and can digitally sign results to help detect tampering before a strategy moves into live trading.
Run demanding neural-trader backtests, parameter comparisons, and model training in an Anthropic Managed Agent cloud container. Estimate costs, test inputs on a small run, collect performance metrics and CSV files, and verify signed results before storage.
Calculate portfolio weights from a covariance matrix, which describes how asset returns move together, and expected returns using Conjugate Gradient. Save the weights with the solver used, numerical error, and timing so the calculation can be checked later.
Convert chat messages into template strings for local model runtimes such as Ollama or llama.cpp. Optionally route queries among a few registered intents using embeddings, numerical representations of meaning, without calling a model.
Explain a trading model's prediction by ranking the input features that contributed to its signal. Returns a Markdown table and saves an attribution record with calculation settings, adding a digital signature when a signing key is available.
Generate RuVLLM settings for local language model inference and create MicroLoRA or SONA instances. Adjust their small internal weights using quality scores as feedback, with instances held in server memory for subsequent adaptation calls.
Compare models available through the HuggingFace router with chat-ui configuration, research missing models, and write short English descriptions of their architecture and capabilities. Add entries to both production and development configuration files and commit the changes.
Manage stored agent memories for backups, project transfers, and team knowledge sharing. Work with RVF, a portable memory format, to export and import records, upgrade formats, inspect storage statistics, and transfer patterns between local projects.
Convert text, code, or documents into 384-dimensional vectors, numerical representations used for meaning-based search, similarity comparison, and clustering. Save vectors to files, check their dimensions and magnitude, and optionally record metadata in AgentDB.
Classify market conditions as bullish, bearish, ranging, or volatile to guide strategy selection and historical testing. Distinguish synthetic demonstrations from analysis of actual price data, with confidence levels and historical comparisons where available.
Choose an agent and model tier using the task description and learned patterns. The skill can explain its recommendation, use compiler-based edits for supported simple changes, and record task outcomes to inform future routing decisions.
Define reusable reasoning and decision patterns for agents, embed them in executable workflows, and review their performance. Share successful approaches, including conditional decision trees, stepwise problem solving, analogies, and evaluation from multiple perspectives.
Capture token usage from Claude Code session logs, calculate USD costs using the supplied pricing reference, and save structured records to AgentDB. Review totals by model and tier, and provide data for cost reports.
Experiment with the policies that govern an AI agent, such as planning, memory, and tool use. Test one policy change per variant in a sandbox, an isolated environment, and promote only candidates with measured improvements.