How to understand the workflow feature & Will agents follow my repo's existing agent rules?

[Wenn5]

Hi Paca team,

First of all, thank you for your brilliant work! Paca is incredible, and I’m certain it will gain huge popularity in the days ahead! 🍺

I fully recognize that Paca is a newly launched project and your team is hard at work improving it, so gaps in documentation are understandable. I’m excited to experiment with its latest features and have outlined several questions below:

---

I'm trying to understand the intended usage model of the Workflow feature and how it fits into a real-world AI software development workflow.

From what I understand, a Workflow is attached to a specific task, and the Workflow itself contains multiple tasks/nodes that define the execution flow.

My understanding is that Workflow is not a global reusable software dev "lifecycle template", but rather an execution plan/orchestration graph for a specific complex task. which means I need to create "workflows" for EACH task (not reusable) Is this correct?

---

My second question is about repository-level agent instructions.

In my projects, I already maintain repository instructions such as:

• AGENTS.md
• CLAUDE.md
• architecture documentation
• coding conventions
• testing requirements

These files define how an AI coding agent should work inside the repository.

When using Paca's built-in agents:

1. Will the agents automatically read and follow existing repository instructions such as AGENTS.md / CLAUDE.md?

2. Can Workflow steps enforce requirements such as:

   • read architecture documents before coding
   • produce a design document before implementation
   • run tests before completion
   • update documentation after code changes

My goal is to use Paca as an AI-native software development process manager, where agents follow a controlled SDLC process instead of directly jumping from an issue description to code changes.

Thanks!

Regards,
JZ

[pikann]

Hi @Wenn5 ,

Thank you so much for your kind words and support! You’ve got a great understanding of Paca's core concepts. Here are the quick answers to your questions:

1. Workflow Model
   You are spot on! Workflows are task-specific orchestration graphs designed to automate assignments (for both humans and AI), making it easier to plan and follow big features.

Quick tip: You don't have to build them manually! You can use Paca’s AI Chat to automatically generate and plan workflows for you.

2. Repository Instructions (AGENTS.md, CLAUDE.md, etc.)
   Yes, absolutely! Our agents will automatically read these files. Paca also has an SDD documentation system that the AI prioritizes. Currently, the AI is designed to read the documentation before starting on any task.

3. Enforcing SDLC Steps
   You can definitely do this! To help you enforce a controlled process (like designing and testing before coding), Paca allows you to customize system prompts and the skills list to tailor the workflow behavior exactly to your needs.

Thanks again for the awesome feedback, and we'd love to hear how your experiments go!