Skill UI

A sophisticated, two-stage system designed to capture and internalize user corrections and learnings during ongoing conversations. It automatically detects patterns (e.g., "use X instead of Y") and queues them. Users can manually review and apply these learnings to a persistent knowledge base (CLAUDE.md), ensuring the AI continuously improves its knowledge and adherence to specific user context for future sessions.

This guide details the production-grade reference architecture for running complex GPU compute workflows on Vast.ai. It covers the foundational three-tier pattern—Orchestrator, GPU Workers, and Artifact Storage—providing blueprints for job queue management, fault-tolerant training pipelines, and robust checkpointing. Use this when designing scalable and reliable ML/AI training systems.

SSH job queue for multi-seed/multi-config ML experiments with OOM-aware retry, stale-screen cleanup, and wave-transition race prevention. Use for batch experiments, grid runs, multi-seed sweeps, or auto-chain experiments when /run-experiment cannot handle 10+ jobs that need orchestration.

Implements robust rate limiting, backoff logic, and quota management for the Anthropic Claude API. It handles 429 errors and optimizes request throughput by managing Requests Per Minute (RPM), Input Tokens Per Minute (ITPM), and Output Tokens Per Minute (OTPM). This skill provides advanced strategies including SDK retries, custom header awareness, queue-based throttling, and batch processing for stable, high-throughput production applications.

This skill provides comprehensive solutions for managing rate limits and throttling when interacting with the AssemblyAI Speech-to-Text API. It implements advanced patterns like exponential backoff with jitter, queue-based concurrency control, and resilient streaming retry logic. Use it when building high-throughput, reliable applications that process large volumes of audio data.

A comprehensive guide to production-ready patterns for ElevenLabs' TypeScript and Python SDKs. This resource covers crucial best practices including singleton client implementation, type-safe TTS service wrappers, robust error classification (rate limiting, quota), and concurrency management using request queues. Essential for building scalable and reliable voice AI applications.

This skill provides utilities to manage and drain the deliberation queue (brain_frames) of the ACC system. It allows users to list waiting frames, retrieve necessary context, perform deliberation on open tasks, and submit continuations using `acc_act(runtime="continue")`. It is essential for checkpointed work saved by headless runs, ensuring continuous task resolution and calibration against predicted outcomes in complex AI workflows.

This skill manages the acc deliberation queue by draining open or waiting brain_frames checkpointed during headless runs. It allows listing frames, reading context, deliberating, and submitting proposals via acc_act with runtime continue. It ensures safe resubmission and calibrates the Work Model using PREDICT lines.

This skill enables horizontal scaling of Large Language Model inference clusters on Kubernetes. It utilizes KEDA for event-driven autoscaling based on custom GPU metrics like request queues and cache usage. The guide covers deploying vLLM on GPU nodes, configuring Prometheus for monitoring, implementing spot instance strategies for cost efficiency, and setting up queue-based scaling for batch jobs. It ensures optimal resource utilization and cost management for high-traffic AI workloads.

Execute exactly one explicitly assigned YYLO Ledger task through the Ralph loop to a validated, queued commit using the yy CLI. Handles task admission, preflight and finish while never pushing, deploying, merging or releasing.