extend-agent

Implements user-requested changes to AgentOS agents and verifies them against the live container.

Updated Sep 9, 2026
One-click install
npx skills add https://github.com/martin-cousseau/refund-arena --skill extend-agent-martin-cousseau
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: extend-agent
Source: https://github.com/martin-cousseau/refund-arena/tree/main/backend/.agents/skills/extend-agent
Command: npx skills add https://github.com/martin-cousseau/refund-arena --skill extend-agent-martin-cousseau

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Modifying a running Agno AgentOS agent involves many failure points: wrong import paths, shadowing Studio-built components with code files, broken container syncs, and unverified edits. This Skill provides a structured loop that grounds each change in documentation, applies it safely, and smoke-tests it against the live backend. ## Core Features & Use Cases - Guided change workflow: Add tools (MCP servers, toolkits, function tools), capabilities (knowledge bases, learning/memory, sub-agents, scheduled tasks), refine instructions, or fix bugs on a named agent. - Registry growth: Declare new building blocks in the platform registry so components built at runtime gain new capabilities, with collision checks on tool names. - Live verification: Restart the container, confirm edits reached it, and smoke-test via curl against the running agent, inspecting which tools fired in logs. - Use Case: A user asks to add an Exa search tool to their support agent. The Skill checks the agno-docs MCP for the correct import, verifies it inside the container, edits the agent file, restarts, and confirms the tool fires on a test prompt. ## Quick Start Ask the assistant to add a web search tool to your agent using the extend-agent workflow and verify it works on the live server.

Frequently Asked Questions about extend-agent

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add a tool to an existing Agno agent?▼

Identify the agent slug, search the agno-docs MCP for the toolkit's import path and constructor args, verify the import inside the container with docker exec, then edit the agent file and restart. Smoke-test with a curl POST to the agent's runs endpoint and check logs for the tool call.

How do I add memory or learning to an Agno agent?▼

Wire the shared learning object via the learning parameter on the agent. Do not use enable_user_memories, which is not an Agent parameter, and keep enable_agentic_memory off beside a LearningMachine. Restart the container and verify with a targeted prompt.

Can I edit an agent built in AgentOS Studio with code?▼

No. Components with is_component=true have no source file and must be changed through Platform Builder tools like edit_agent and publish_component. Creating a code file under the same id silently shadows the built component because code wins on resolution.

Why did my agent edit not take effect after restart?▼

The container may not be bound to your current checkout, or the edit never reached the container filesystem. Verify the bind mount with docker inspect, then grep inside the container for a unique substring from your edit; a count of zero means the edit did not land.

When should I use extend-agent versus improve-agent?▼

Use extend-agent when the user names a concrete change such as adding a tool, capability, or bug fix. Use improve-agent for autonomous hardening when there is no specific change in mind; it is typically run after extend-agent finishes.