Skill UI

This skill provides comprehensive web browser automation capabilities for agents. It utilizes AI-optimized snapshots, reducing context size by 93% by using element references instead of full DOM. Functions include navigation, interactive element extraction, form filling, data scraping, and state management across multiple isolated sessions. Ideal for complex web testing and data extraction workflows.

A robust framework for orchestrating complex, multi-step coding tasks using multiple AI agents. It functions by first planning the entire project into a mission dependency graph (DAG), then dispatching specialized agents to isolated Git worktrees for parallel development. The system automatically monitors progress, handles merges, and generates structured, comprehensive reports detailing changes, completed tasks, and next steps.

A formal framework designed for evaluating AI-assisted code generation sessions. It implements Development Oriented Evals (EDD) principles, allowing users to define capability and regression criteria. The system supports multiple evaluation types—code-based, model-based, and human review—and tracks reliability using advanced metrics like pass@k, ensuring robust quality control for AI agents.

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).

A comprehensive, end-to-end playbook designed for software engineers and AI agents. It provides a structured methodology to safely introduce machine learning capabilities into existing, non-ML codebases. The process covers critical stages: problem framing, data readiness, architectural decoupling (using dedicated APIs/services), baseline model implementation, and final handoff to MLOps practices (CI/CD, monitoring).

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.

This tool initializes a structured multi-agent swarm with built-in anti-drift configuration. It is designed for orchestrating complex, multi-file tasks that require coordinated efforts from three or more specialized AI agents, such as feature implementation, cross-module refactoring, or comprehensive security audits. It supports various topologies (hierarchical, mesh, etc.) to ensure robust and manageable parallel work execution across large teams.

Tree Ring Memory is a local-first, framework-agnostic layer for managing durable knowledge in AI agents. It provides structured methods for recalling past decisions, recording evidence, auditing project state, and deliberately forgetting information, ensuring the core project memory remains clean and actionable without relying on raw conversation transcripts.

This tool suite provides functionality to interact with a community gallery for WebAssembly (WASM) agents. Users can browse available agents, search for specific functionalities, install pre-built agents, publish their own creations, and export agents for packaging. It streamlines the process of sharing and utilizing modular WASM components.

This guide outlines best practices for ensuring rigorous reproducibility of complex interaction claims, especially in game theory and multi-agent systems (MAS). It details required evidence for proofs, experimental setups (agents, opponents, seeds, compute), and consistency checks between the paper's claims and the provided artifacts. Essential for high-stakes academic submissions like AAMAS.

Systematically transforms high-level ideas or goals into a detailed, structured, and executable taskbook designed for AI agents. This process involves deep research, validation against existing codebases, adherence to strict formatting rules, and generating a single, copy-pastable block of instructions for immediate agent execution. It distinguishes between execution and exploration tasks, ensuring all steps are verifiable and anti-cheating measures are included.

This skill provides comprehensive command-line interface (CLI) support for OpenViking, enabling agents to manage external resources, perform filesystem operations, and execute data tasks. It covers adding resources, importing/exporting data, browsing directories, reading/writing files, and advanced searching (find, grep, search). Use cases include data pipeline setup, knowledge graph manipulation, and system maintenance.