The Agent Skills Directory

Local Command Execution: The skill utilizes shell commands through the Bash tool to perform development tasks.
Evidence: Use of hamilton build for DAG validation and monkeytype run for type inference.
Context: These commands are used to automate the validation of the user's graph structure and to improve code quality through automated type annotations. They are restricted to specific binaries (python, hamilton, pytest) in the environment configuration.
Code Generation Surface: The skill guides the agent to generate Python implementations based on user-provided natural language requirements or DOT graph descriptions.
Evidence: Steps 1 and 2 describe converting requirements into DOT language and then into Python function signatures.
Context: As a development workflow, the agent generates code that the developer is intended to review. While this represents a surface for indirect instructions, it is governed by the agent's primary safety protocols for code generation.
Dynamic Execution for Type Inference: The workflow suggests using MonkeyType to infer types at runtime.
Evidence: monkeytype run -m pytest tests/.
Context: This is a standard practice in Python development to generate accurate type hints based on actual runtime behavior. It executes the local test suite to observe data types.

hamilton-dev-workflow