Feature description
I propose implementing a stateful conversational agent designed to interactively generate project documentation in a strict hierarchy:
Specification $\rightarrow$ Plan $\rightarrow$ Tasks
Leveraging the conversational state patterns from hiking_agent.py and the file system tool capabilities from file_explorer_agent.py, this feature will allow the agent to:
- Maintain Hierarchical Context:
- The agent guides the user through three distinct stages.
- Spec: High-level project description and goals.
- Plan: Breakdown of the Spec into specific features and architectural details.Tasks: Granular actionable items derived directly from the Plan.
- Constraint: The agent maintains the conversation history within the current session to ensure consistency (e.g., Tasks generated must align with the previously agreed-upon Plan).
- Implement Approval-Based Progression:
- The agent remains in the current stage (looping and refining) until the user explicitly approves the content.
- Transitioning to the next level (e.g., from Spec to Plan) "locks" the previous context as the foundation for the next step.
- Optional File Persistence:
- Upon approval of a stage, the agent can optionally utilize a file-writing tool (based on file_explorer_agent) to export the content into formatted Markdown (.md) files.
- This allows the user to choose between a purely in-memory brainstorming session or generating tangible documentation artifacts.
Motivation
I started my journey with Vibe Coding this month. After diving into Spec-Driven Development and attending the first day of our internal training on 'Security & Privacy in the Age of Vibe Coding,' I realized how crucial it is—both for AI agents and for us—to define precise solutions. When diving deep into technical details, it’s easy to lose sight of the project's main goal. This tutorial will show developers how to craft the right instructions through effective prompting, covering everything from the initial concept to final tasks.
Additional context
As a result, we can prepare these types of spec files for Ragbits projects, which can then be used by everyone involved internally.
Feature description
I propose implementing a stateful conversational agent designed to interactively generate project documentation in a strict hierarchy:
Specification$\rightarrow$ Plan $\rightarrow$ Tasks
Leveraging the conversational state patterns from hiking_agent.py and the file system tool capabilities from file_explorer_agent.py, this feature will allow the agent to:
Motivation
I started my journey with Vibe Coding this month. After diving into Spec-Driven Development and attending the first day of our internal training on 'Security & Privacy in the Age of Vibe Coding,' I realized how crucial it is—both for AI agents and for us—to define precise solutions. When diving deep into technical details, it’s easy to lose sight of the project's main goal. This tutorial will show developers how to craft the right instructions through effective prompting, covering everything from the initial concept to final tasks.
Additional context
As a result, we can prepare these types of spec files for Ragbits projects, which can then be used by everyone involved internally.