Skip to content

Conversation

@pinkeshmars
Copy link
Collaborator

Description

Add guidelines for effective prompting

Linear ticket and magic word Fixes DEVR-1097

Type of change

  • Typo fix
  • New feature
  • Enhancement to current docs
  • Removed outdated references
  • Update assets

@pinkeshmars pinkeshmars requested a review from PoojaB26 January 2, 2026 05:39
@PoojaB26
Copy link
Collaborator

this should be under a new section called concepts and not in agent panel. @pinkeshmars

@pinkeshmars
Copy link
Collaborator Author

this should be under a new section called concepts and not in agent panel. @pinkeshmars

ok, and where should the new 'Concepts' section be placed? Should it be at Root Docs → Concepts (after the Debugging section), or nested inside the Workspace section?

@PoojaB26
Copy link
Collaborator

PoojaB26 commented Jan 21, 2026

ok, and where should the new 'Concepts' section be placed? Should it be at Root Docs → Concepts (after the Debugging section), or nested inside the Workspace section?

Yes, at the root, after Debugging section.

@pinkeshmars
Copy link
Collaborator Author

@PoojaB26 the content is now moved to the 'Concepts' section.

Copy link
Collaborator

@PoojaB26 PoojaB26 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Need few changes in some sections + can we also add John's video on prompting as a reference? https://youtu.be/eFtTpmtIujo?si=WbwzQhW3hTODTrgO


Dreamflow’s Agent is most effective when you treat it like a **collaborative developer**, not a one-shot code generator. Clear prompts, resolved ambiguity, and small scopes lead to better code, lower credit usage, and more predictable results.

Many coding prompts requests contain hidden ambiguity. A prompt like “Create this UI” or “Add favorites feature” can imply different data sources, navigation patterns, persistence, UX behavior, and architecture. If you do not specify these, the Agent has to guess, which can lead to extra code, mismatched behavior, and more iterations.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

also mention - leading to more credit usage and wastage


With strong rules in place, your prompts can stay short while still producing consistent, production-ready output.

### The Agent as a Junior Developer
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

i dont think we need this section, its already too text heavy,


The Agent is a junior developer who is fast and capable, but still needs clear direction and review. When you prompt with strong constraints and iterate step by step, you get a codebase that scales and stays maintainable.

## FAQ
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The FAQs are already available in Agent Panel doc, why repeat here?


Dreamflow's AI Agent is strongest when you break work into small, reviewable steps.

Prefer prompts like:
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

the prompts still feel low effort or less detailed for the agent. instead this section should talk about adding context to widget features of the app, for example, adding a widget to agent for exactly pointing which widget is being referred to and then give an example of a prompt and then also use screenshot feature for references to enrich the context further.
if possible a quick arcade would be great since this page is too text heavy. lets re-write this section keeping context and detailed task (not short) in mind and maybe update the title accordingly?

Co-authored-by: Pooja Bhaumik <pbhaumik26@gmail.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants