families / agent-tooling
agent-tooling
agent-orchestration-frameworks
A collection of specialized tools and frameworks for building, monitoring, and managing autonomous AI agent workflows and infrastructure.
These skills differ by their specific role in the agent lifecycle, ranging from core execution loops and infrastructure management to specialized monitoring and utility integrations.
- Set up a new autonomous agent loop using the new-loop framework to handle recurring data processing tasks.
- Configure the aeon-monitor-runners to track the performance and status of multiple distributed agent instances.
ai-model-integration
A collection of tools and frameworks for routing, optimizing, and integrating various AI models into agent workflows.
These skills differ by their specific focus on model selection, framework implementation, or performance optimization within the AI agent stack.
- Configure the ai-model-router to direct queries to the most cost-effective model based on complexity.
- Use pydantic-ai to define structured outputs for an LLM agent and validate them against a schema.
alfworld-object-manipulation
A set of skills for locating, picking, heating, cooling, cleaning, and disposing of objects within the ALFWorld environment.
These skills focus on direct interaction with and state modification of specific objects, distinct from navigation, goal interpretation, or general environment scanning.
- Find a dirty mug, clean it, and put it in the sink.
- Locate a frozen apple, heat it in the microwave, and place it on the table.
autonomous-agent-orchestration
A family of skills for configuring, steering, and executing self-directed AI agents that operate with minimal human intervention.
These skills differ in their specific implementation frameworks and levels of autonomy, ranging from general-purpose autonomous loops to specialized operational or custom steering configurations.
- Configure an autonomous agent to monitor server logs and automatically deploy fixes without approval.
- Set up a custom steering loop for an AI agent to iteratively refine code until tests pass.
browser-automation
A collection of tools and agents for automating web browsing, data extraction, and interaction with web applications.
These skills differ in their underlying infrastructure, such as using local headless browsers, cloud-based session management, or specific AI-driven agent frameworks for navigation.
- Scrape product prices from an e-commerce site and save them to a CSV file.
- Automate the login process for a dashboard and take a screenshot of the main page.
accesslint
A suite of tools for auditing, scanning, and diffing accessibility compliance in codebases.
These skills differ by their specific operation on accessibility data: scanning for initial detection, auditing for comprehensive reporting, and diffing for change analysis.
- Run a full accessibility audit on the current project and report violations.
- Compare the accessibility status of the main branch against the feature branch to identify regressions.
agent-cli-tooling
A collection of command-line interfaces designed for interacting with AI agents, managing their configurations, and routing tasks within an agent-native environment.
These skills differ by their specific target platforms or functions, such as general agent initialization versus specific vendor CLIs like blucli or zai-cli, and task routing capabilities.
- Initialize a new agent project using the ai-native-cli and configure its dependencies.
- Route a specific user query to the appropriate backend agent using the lark-cli-router.
context-lifecycle-management
A family of skills dedicated to the creation, optimization, compression, restoration, and maintenance of agent context windows to ensure efficient and accurate long-running sessions.
These skills differ by their specific focus within the context lifecycle, ranging from foundational understanding and optimization strategies to active compression, restoration of prior states, and diagnosis of performance failures.
- Compress the current session history into a durable handoff summary to free up token space for the next task.
- Diagnose why the agent is ignoring instructions in the middle of the prompt and apply mitigation strategies for context degradation.
delegate-coding-task
Hand off a specific coding implementation task to a designated external CLI agent to run in the background, then review and merge the resulting diff.
The skills differ by the specific external CLI tool (e.g., Codex, Cline, Claude, Cursor) used to perform the background implementation.
- Delegate the implementation of the new authentication middleware to the Codex CLI and review the changes when it finishes.
- Have the Cline agent refactor the database connection pool in the background so I can review the diff later.
project-skill-audit
Analyze a project's past Codex sessions, memory files, and existing local skills to recommend the highest-value skills to create or update.
This skill focuses on analyzing existing project history and memory to recommend new or updated skills, whereas other skills in the list are for specific implementation tasks like setting up Playwright, managing second-brain notes, or executing skillshare workflows.
- What skills should I create for this project based on my recent usage?
- Audit my current local skills and suggest which ones need updating.
gmail-management
A suite of tools for interacting with Gmail to read, send, reply, triage, and monitor email communications.
These skills differ by their specific interaction mode with the mailbox, ranging from passive monitoring and reading to active composition, replying, and automated triage.
- Read the latest unread emails and summarize the key points.
- Reply to the most recent email from my manager with a brief update.
llm-integration
A collection of skills for selecting, configuring, and architecting Large Language Model integrations within software projects.
These skills differ by focusing on specific provider APIs, framework orchestration, architectural decision-making, or performance optimization techniques like caching.
- Which free LLM API should I use for a high-volume summarization task?
- How do I structure a multi-stage agent pipeline using LangChain and Qwen?
mcp-integration
A collection of skills for building, wrapping, benchmarking, and utilizing Model Context Protocol (MCP) servers and tools.
These skills differ by their specific role in the MCP lifecycle, ranging from general server construction and platform-specific integrations to security wrapping and performance evaluation.
- Create a read-only wrapper around an existing MCP server to ensure safe data access.
- Benchmark the performance of the Taobao MCP server against specific latency metrics.
amemo-memory-management
A suite of tools for initializing, saving, retrieving, and managing persistent memory artifacts, tasks, and data within the Amemo agent framework.
These skills are distinguished by their specific interaction with the Amemo system's state, handling distinct entity types such as memos, tasks, data, and mates, as opposed to generic memory operations.
- Save the current project context as a memo and then retrieve the last saved data point.
- Initialize a new mate session and send a specific task to the agent for processing.
model-management
A set of skills for configuring, switching, routing, and migrating between different AI models and their associated prompts.
These skills differ in their specific focus, ranging from real-time switching and routing logic to static configuration, prompt packaging, and long-term migration planning.
- Set up automatic model switching based on task complexity using the smart-model-switcher.
- Create a migration plan to move our current prompts from one model provider to another.
multi-agent-orchestration
A collection of skills for designing, coordinating, and optimizing systems involving multiple interacting AI agents.
These skills differ in their specific orchestration patterns, such as council-based planning, swarm coordination, or performance optimization, rather than single-agent tasks.
- Design a supervisor-based architecture for a customer support system that isolates context between agents.
- Optimize the workload distribution and cost efficiency of an existing multi-agent pipeline.
n8n-workflow-development
A comprehensive set of skills for designing, configuring, and debugging n8n workflows, including AI agents, expressions, and MCP tool integration.
These skills differ by their specific focus within the n8n ecosystem, ranging from low-level syntax validation and node configuration to high-level AI agent design and MCP server interaction.
- Configure an AI Agent node with a custom code tool and validate the expression syntax for data mapping.
- Use the n8n-mcp server to search for a specific node, configure its properties, and organize the resulting workflow into a folder.
plugin-lifecycle-management
A set of skills for creating, structuring, building, and installing software plugins within the OpenClaw ecosystem.
These skills differ by their specific stage in the plugin lifecycle, ranging from initial creation and structural definition to final building and installation.
- Generate the directory structure and manifest for a new OpenClaw plugin.
- Install a third-party plugin from a public repository into the local agent environment.
proactive-agent
A family of agent configurations designed to initiate actions, monitor environments, and anticipate user needs without explicit prompting.
These skills differ primarily in their specific implementation details, such as the underlying framework, complexity level (e.g., 'lite' vs. full), or the specific proactive behaviors they prioritize (e.g., monitoring vs. drafting).
- Configure an agent to automatically summarize new emails and draft replies based on my calendar availability.
- Set up a proactive monitor that alerts me when server resource usage exceeds 80% and suggests scaling options.
llm-prompt-engineering
A collection of tools and techniques for designing, optimizing, evaluating, and managing prompts for large language models.
These skills differ in their specific focus, ranging from high-level strategic pattern design and expert consultation to automated optimization, code-based frameworks like DSPy, and specific utility tasks like saving or summarizing prompt portfolios.
- Refactor this system prompt to enforce strict JSON output and improve instruction clarity.
- Generate a suite of test cases and an evaluation rubric to measure the performance of a new customer support prompt.
scienceworld-container-management
A set of skills for inspecting, retrieving, relocating, and transferring items within containers in the ScienceWorld environment.
These skills specifically focus on the interaction with and manipulation of items within containers, distinguishing them from skills related to circuit building, animal identification, or general action resolution.
- Check what is inside the refrigerator and then move the apple to the table.
- Retrieve the key from the drawer and transfer it to the pocket.
agent-self-optimization
A collection of skills enabling agents to autonomously refine their own logic, manage context, and schedule tasks to enhance performance over time.
These skills differ in their specific mechanisms for autonomy, ranging from explicit self-improvement loops and context management to scheduling and thinking frameworks.
- Configure a self-improvement loop that analyzes past failures and updates the agent's prompt strategy.
- Implement a self-managed context window that prunes irrelevant history to maintain focus during long sessions.
session-lifecycle-management
A set of tools for auditing, summarizing, recovering, and transferring context across AI agent work sessions.
These skills differ by their specific role in the session lifecycle, ranging from post-hoc auditing and log retrieval to active context preservation and recovery for future sessions.
- Audit the last three coding sessions for token usage and tool failures using agenttrace.
- Write a handoff summary for the current session so a new agent can pick up the work where I left off.
agent-skill-lifecycle
A collection of tools for creating, evaluating, repairing, and managing agent skills within a registry or local environment.
These skills differ by their specific role in the skill lifecycle, ranging from creation and installation to quality assurance, repair, and registry management.
- Run the evaluation suite for the 'aeon' skill and generate a report on its performance.
- Scan the local directory for new skills and index them into the central registry.
task-orchestration
A set of skills for creating, routing, delegating, and executing tasks across autonomous agents and human workflows.
These skills differ in their specific role within the task lifecycle, ranging from initial creation and planning to active routing, delegation, and execution monitoring.
- Route this new bug report to the appropriate AI agent for immediate resolution.
- Create a detailed task plan for the upcoming sprint and delegate subtasks to the team.
token-optimization
A family of skills designed to minimize token consumption and improve efficiency in agent interactions.
These skills differ in their specific implementation strategies for reducing token overhead, ranging from strict zero-token protocols to heuristic-based dieting and optimization algorithms.
- Compress the current context window to reduce token usage.
- Analyze the prompt history to identify and remove redundant tokens.
agent-tool-lifecycle
A set of skills for selecting, building, routing, and securing tools for AI agents.
These skills differ by their specific role in the tool lifecycle, ranging from initial selection and construction to runtime routing and security enforcement.
- Build a custom tool for a personal agent and configure its routing logic.
- Select the best existing tool for a task and apply security guards to its usage.
voice-agent-development
A collection of skills for designing, building, and operating voice-based AI agents using various frameworks and models.
These skills differ by their specific focus on either high-level design, specific framework implementation (like Pipecat or Nemotron), or operational aspects of voice agents.
- Design the conversation flow and system prompts for a customer support voice agent.
- Build a real-time voice agent using the Pipecat framework with a specific LLM backend.
web-browsing-and-search
A collection of skills for interacting with the web, ranging from lightweight API-based search and content fetching to full browser automation and session management.
Skills differ by their interaction method, with some using lightweight APIs for search and fetching (like brave-search or fetch) while others utilize full browser automation for complex interactions and session persistence (like use-browser or safe-browser).
- Search the web for the latest documentation on React hooks and extract the key points.
- Log into the internal dashboard, navigate to the reports section, and download the monthly summary PDF.
agent-workflow-orchestration
A set of skills for designing, executing, and managing complex multi-step agent workflows and automation pipelines.
These skills differ by their specific role in the workflow lifecycle, ranging from high-level design and automation setup to granular control commands like pausing, resuming, and securing specific phases.
- Design a secure multi-phase workflow for an autonomous builder that includes pause and resume points.
- Create a taskflow automation that wraps up a spike and moves to the next workstream using openclaw.
general
Skills filed under general.
The skills differ by name and by the task each description states.
- Use find
- Apply find to a matching task