Agent Skills Hub

debug-distributed

Debugging guide for AReaL distributed training issues, including hangs, NCCL errors, OOM, and numerical consistency in FSDP2/TP/CP/EP.

research-pipeline

End-to-end autonomous research agent: from idea generation and literature review to experiment execution, adversarial review loops, and paper writing.

shift-right-testing

Production-grade testing strategy implementing feature flags, canary releases, synthetic monitoring, and chaos engineering for continuous reliability in live environments.

marketing-psychology

Apply behavioral science, mental models, and psychological principles to marketing strategy, copywriting, and decision-making.

yc-advisor

Access Y Combinator’s library of 443+ startup resources for expert advice on fundraising, co-founders, product development, growth, and scaling your business.

milady-development

Self-modify your Milady agent by managing plugins. Edit code, rebuild, and restart the runtime to develop new capabilities or improve agent workflows locally.

MLOps Industrialization

A framework to transform experimental ML prototypes into robust, production-ready Python packages using src layout, hybrid architecture, and strict configuration management.

margin-management

Monitor and manage margin-living strategy by tracking balances, interest costs, and coverage ratios. Provides automated scaling recommendations and safety alerts based on portfolio-to-margin thresholds.

evaluating-code-models

Evaluate code generation models using BigCode Evaluation Harness. Benchmarks include HumanEval, MBPP, and MultiPL-E with pass@k metrics for multi-language coding models.

meteorology-driver-classification

Classify and group meteorological and environmental variables into specific driver categories for consistent attribution analysis and environmental modeling.

literature-engineer

Evidence-first literature collector for automated research pipelines. Scales paper pools to 1200+ with metadata normalization, provenance tracking, and multi-source ingestion.

guard

Epistemic safety analysis for JSON data in prompts to prevent LLM hallucinations and reasoning errors when handling incomplete or large-scale datasets.