Add development prompts with testing structure #23
Conversation
- Add prompts.py module with generation functions for tools, agents, sessions, models, and multi-agent patterns - Create Jinja2 templates for all development scenarios - Add extensive test coverage for prompt generation - Integrate new prompt endpoints into MCP server - Include regex filtering and content fetching utilities
- Update dev dependencies to latest versions (pytest 8.0+, pytest-asyncio 0.23+) - Add pytest-xdist for parallel test execution - Add pytest-mock and mypy for better testing and type checking - Configure hatch-test environment with Python 3.10-3.13 matrix - Add comprehensive test, lint, and prepare scripts - Configure coverage reporting (HTML and XML output) - Add pytest configuration with test paths - Add mypy type checking configuration - Include tests directory in ruff linting
- Add test-lint.yml workflow for unit tests and linting - Test matrix: Python 3.10-3.13 on Linux, Windows, macOS - Comprehensive linting with ruff - Add pr-and-push.yml for automated PR and main branch checks - Concurrency control to prevent duplicate runs - Automatic test and lint on PR events - Add pypi-publish-on-release.yml for automated PyPI deployment - Uses PyPI Trusted Publisher for secure release - Validates version format before publishing - Add PULL_REQUEST_TEMPLATE.md for standardized PR submissions - Update dependabot.yml with pip and GitHub Actions monitoring
There was a problem hiding this comment.
In the the PR description, can you expand on what the user-experience/job-to-be-done is?
Coming into this, I'm not sure I understand what we're doing and what benefit this is.
There was a problem hiding this comment.
The idea is to empower coding assistants to develop with Strands Agents. Essentially the prompts should provide the necessary up to date template info for developing with Strands.
It should make development easier, so I can just start prototyping with a couple lines.
We have agent-builder for a similar reason, but people tend to stick to their tools (Cursor, Cline, Q CLI, etc.). So this is an easy way to improve the DevX
There was a problem hiding this comment.
Can you document the experience? My understanding is that coding assistants typically have MCP prompts show up as explicit commands you can invoke. Does that match how you're thinking about this feature?
For example, in Q CLI:
@strands_agent_development "my use case" "my examples" "my agent guidelines" "my tools_required" "my model preferences"
In Cursor: /strands_agent_development and then an input box is supposed to pop up, I think?
4 participants
🎯 Overview
This PR introduces a comprehensive prompt generation system for the Strands MCP Server, enabling intelligent context-aware prompts for tool development, agent creation, session management, model configuration, and multi-agent orchestration. Additionally, it establishes a robust testing infrastructure and CI/CD pipeline to ensure code quality and maintainability.
The goal of these prompts is to give necessary context and guidelines to LLMs, to make getting started much easier. The prompts are dynamically generated using up to date documentation.
🎯 User Experience
What Are MCP Prompts?
MCP Prompts are specialized templates that enhance AI coding assistants (Cursor, Q CLI, Cline) with comprehensive Strands context. When developers invoke a prompt, their AI assistant receives up-to-date documentation, best practices, and structured guidance for building with Strands Agents.
How Developers Use Them
@strands_agent_development "Create a document analyzer"/strands_agent_developmentin Composer/Agent modeThis brings the agent-builder experience directly into developers' preferred tools, eliminating context switching and making Strands development faster and more intuitive.
✨ Key Features
1. Prompt Generation System
prompts.pywith Jinja2 templates2. Testing Infrastructure
3. CI/CD Pipeline
🧪 Testing
✅ Checklist