Skill UI

This skill provides a systematic workflow for diagnosing failures in GAIA benchmark runs. Users can analyze task traces, classify failure modes (e.g., Tool Gap, Reasoning Miss, Extraction Bug), identify root causes, and apply targeted fixes to improve AI agent performance. It is essential for optimizing LLM benchmarks and finding performance regressions.

A comprehensive multi-agent feedback loop inspired by Generative Adversarial Networks (GANs), designed to build complex, production-ready applications. It separates generation from evaluation, using a 'ruthlessly strict' evaluator (QA) to drive continuous iteration. This process ensures high design quality, robust functionality, and originality, overcoming the limitations of single-agent self-critique. Ideal for full-stack feature development and complex product specification implementation.

This skill runs an automated evolution process designed to systematically optimize the seven core policy surfaces of an AI agent harness (e.g., planner, contextBuilder, reviewer). By mutating these policies and scoring the variants in a sandbox, it identifies crucial performance improvements without requiring expensive foundation model retraining. It is ideal for empirical configuration tuning and establishing optimized baselines.

A powerful tool designed to scaffold sophisticated, custom AI agent harnesses using the metaharness CLI. It supports various templates (e.g., coding, devops, marketing) and AI hosts (e.g., Claude, Codex). Crucially, it operates with maximum safety, defaulting to a dry-run mode and refusing to write to the calling repository root, ensuring controlled and predictable development workflows.

A sophisticated system designed to automatically select the optimal AI agent and model tier (e.g., Haiku, Sonnet, Opus) for any non-trivial task. It replaces manual decision-making by integrating learned patterns, predictive modeling, and a robust 3-tier routing mechanism. The process is fully auditable, providing a detailed routing rationale to ensure transparency and reliable execution.

This tool is designed to register a Cognitum Seed IoT device, establishing a secure agent bridge and fetching the device's identity. It resolves the specified or default endpoint for connectivity and handles initial trust level assignment. It is crucial for the initial setup, pairing, and status checking of IoT hardware deployments.

This tool is designed to analyze source code by extracting core entities (classes, functions, modules, types) and mapping their complex relationships (imports, inheritance, calls, dependencies). It constructs a comprehensive Knowledge Graph (KG) in AgentDB. This is essential for detailed dependency analysis, understanding large codebase structures, and performing accurate impact assessments.

A comprehensive guide for optimizing latency-critical systems, such as market data feeds, real-time dashboards, and streaming agents. It details structured methods for mapping 'hot paths,' tracking key metrics (p95, p99 latency, throughput, freshness), and implementing architectural improvements like caching and batching to ensure high performance and data freshness under load.

This skill guides the design and rigorous review of goal-oriented agent feedback loops. It helps ensure that complex autonomous multi-step agents are robust, preventing failures such as infinite looping, goal-gaming, or running to completion with incorrect assumptions. It covers defining machine-decidable goals, setting boundaries, and selecting the correct loop type (servo, regulator).

This skill provides access to the Anthropic Claude Managed Agent runtime, which is a cloud-native, persistent environment for running complex, long-running, or asynchronous AI tasks. Unlike local WASM agents, the managed runtime provides a full container, persistent filesystem, and robust state management across multiple turns. It supports the entire agent lifecycle: creation, prompting, status inspection, event logging, and mandatory termination.

A comprehensive, end-to-end methodology designed for software engineers and AI agents to safely and systematically integrate machine learning models into existing non-ML codebases. It guides users through critical phases, including problem framing, data readiness, architectural decoupling (via APIs/services), baseline model implementation, and final MLOps pipeline setup.

This comprehensive skill guides the user through the entire process of creating a deep, multi-layered AI persona (Soul.md). It helps define the identity, core rules, name, and visual prompts for agents designed for the OpenClaw platform. Users can initiate character creation via guided prompts or a random 'gacha' draw, ensuring a cohesive and compelling character profile.