Create, clean, and optimize datasets for LLM fine-tuning. Covers formats (Alpaca, ShareGPT, ChatML), synthetic data generation, quality assessment, and augmentation. Use when preparing data for training.
Helps work with the b00t datum system - TOML-based configuration for AI models, providers, and services. Datums are stored in ~/.dotfiles/_b00t_/ and specify WHICH environment variables are required (not the values). Enables DRY approach by centralizing configuration in Rust, exposed to Python via PyO3.
Implements the b00t environment management pattern: direnv → .envrc → dotenv → .env where datums specify WHICH environment variables are required and .env contains the actual secret VALUES. Ensures automatic environment loading per-project.
Multi-agent data science framework using DS-STAR (Data Science - Structured Thought and Action) architecture. Automates data analysis through collaborative AI agents with multi-model support (Haiku, Sonnet, Opus). Use for exploratory data analysis, automated insights, and iterative data science workflows.
East Execution Engine (e3) - durable dataflow execution for East programs. Use when: (1) Authoring e3 packages with @elaraai/e3 (e3.input, e3.task, e3.package, e3.export), (2) Running e3 CLI commands (e3 init, e3 start, e3 watch, e3 get, e3 set), (3) Working with workspaces and packages, (4) Content-addressable caching and dataflow execution.
Use when writing Ecto code. Contains insights about Contexts, DDD patterns, schemas, changesets, and database gotchas from José Valim.
Containerize Java applications - Dockerfile optimization, JVM settings, security
This skill guides configuring Litestream for continuous SQLite backup in Rails 8+ apps. Use when setting up production backups for SQLite databases (Solid Queue, Solid Cache, Solid Cable).
Add account state (P&L, win rate, drawdown) to RL observations + drawdown penalty in rewards. Trigger when: (1) model needs account awareness, (2) training should penalize drawdowns, (3) upgrading obs_dim 5300→5600.
This skill should be used when analyzing a link to an AI tool and adding it to the awesome-ai-tools readme with proper categorization
Guidelines for selecting appropriate AI model (Sonnet vs Haiku) based on task complexity, ensuring cost efficiency while maintaining quality. Use when assigning work.
Use when starting a fine-tuning project to determine if fine-tuning is needed, or when evaluating whether a base model meets quality thresholds for a specific domain task
Achieve comprehensive baseline (V_meta ≥0.40) in iteration 0 to enable rapid convergence. Use when planning iteration 0 time allocation, domain has established practices to reference, rich historical data exists for immediate quantification, or targeting 3-4 iteration convergence. Provides 4 quality levels (minimal/basic/comprehensive/exceptional), component-by-component V_meta calculation guide, and 3 strategies for comprehensive baseline (leverage prior art, quantify baseline, domain universality analysis). 40-50% iteration reduction when V_meta(s₀) ≥0.40 vs <0.20. Spend 3-4 extra hours in iteration 0, save 3-6 hours overall.
Apply Chiral Narrative Synthesis (CNS) framework for contradiction detection and multi-source analysis using Tinker API for model training. Use when implementing CNS with Tinker for fine-tuning models on contradiction detection, training on SciFact/FEVER datasets, or building multi-agent debate systems for narrative synthesis.
This skill should be used when creating new test lanes for the XML test data generator. A test lane consists of an XSD schema file paired with a meta.yaml configuration file. This skill guides the process of creating both files with proper semantic type mappings, distribution settings, and field overrides. Use when users request new test lanes, want to generate test data configurations, or need help setting up XSD + meta.yaml pairs for the testgen CLI tool.
Master machine learning, data engineering, AI engineering, MLOps, and prompt engineering. Build intelligent systems from data pipelines to production AI applications with LLMs, agents, and modern frameworks.
