Kiro Power for Pydantic AI

There are two ways to install this power in Kiro.

From GitHub:

1. Open the Powers panel (click the Ghosty icon with the lightning bolt 👻⚡ in the sidebar)
2. Click "Install Custom Power" → "Import power from GitHub" and paste the repo URL (https://github.com/JohnCrickett/kiropydanticpower)
3. The power must have a valid POWER.md file in the repository root

From a local folder: For powers you have cloned locally and install from the local path using the "Import power from a folder" option.

Once installed:

To use the power ask Kiro to help create an agent with Pydantic AI, for examples see below.

N.B. This are short examples for illustration, for best result provide a more detailed specification.

Creating a Chatbot

I want to create a chatbot using Pydantic AI and gemini 2.5 flash.

Creating a Structure Output Agent

Build an AI agent using Pydantic. The agent should handle support requests from customers and return structured data.

Creating a Support Agent with Evals

Write an agent to handle support tickets. It should parse user request in unstructured text and generate structured output. 

Use Pydantic AI and I want evals to check it works correctly.

User tickets should have a serveity from 1 (highest) to 5 (lowest).
Tickets should have a title, category, summary and full description. The full description should be what the user provided.

Creating a Tool Calling Agent

Build an AI Assisted coding agent for software engineers using Pydantic AI.

1. It should support tool calls
2. It should support the AGENTS.md file
3. It should support Skills
4. It should support MCP
5. It should support subagents
6. It should support context window management (conversation compression/summarization when approaching limits)
7. It should support a permission system (auto-allow, prompt, deny per tool/action)
8. It should support persistent memory across sessions
9. It should support streaming responses
10. It should support file reading and editing as built-in tools (single-file primitives, multi-file through orchestration)
11. It should support bash/shell command execution
12. It should support codebase search (glob patterns and content grep)
13. It should support git integration (status, diff, commit, PR creation)
14. It should support hooks (user-defined shell commands triggered by agent events)
15. It should support plan mode (architect/plan before implementing)
16. It should support conversation history (resume previous sessions)
17. It should support cost/token usage tracking
18. It should support model selection (switch between models mid-session)
19. It should support a configuration file hierarchy (global, project, local settings)
20. It should support headless/non-interactive mode (CI/automation use cases)
21. It should be model agnostic (support multiple LLM providers)

Creating a Data Analysis Agent

Build a data analysis agent using Pydantic. It should be able to connect to a relational database, i.e. MySQL or SQLite and allow the user to ask questions about the database using a natural language.

Creating a Multi-Agent System

Build a multi-agent CLI system using Pydatantic AI. The agents should work together to fetch documents from a URL, review them for grammatical errors and readability then edit them before re-reviewing. Once approved the documents should be provided to the user for final review.