The Problem
During the mandatory initialization sequence defined in AGENTS_STARTUP.md, Step 4 dictates that the agent should read learn/gettingstarted/DescribingTheUI.md. However, at this specific point in the boot cycle, the agent has not yet fetched the Context Frontier or interacting with the user to determine if the task actually involves frontend UI components. Reading UI guides prematurely when the task might be backend or infrastructure-related wastes cognitive bandwidth and violates context scoping protocols.
The Architectural Reality
The current instructions linearly enforce frontend document processing before the Memory Core determines the strategic task. This must be decoupled. This flaw has caused agents in past sessions to pollute their initial context window with UI assumptions.
The Solution Implementation
- Remove
Step 4: Understand the Two Component Models from the rigid, domain-agnostic session initialization loop in AGENTS_STARTUP.md.
- Reposition the requirement to read
DescribingTheUI.md as a conditional mandate inside the Implementation Loop (Section 4), explicitly stating that the document should only be loaded after the agent confirms the task involves frontend DOM/VDOM components.
This ensures the session initialization remains lean and task-agnostic.
The Problem
During the mandatory initialization sequence defined in
AGENTS_STARTUP.md, Step 4 dictates that the agent should readlearn/gettingstarted/DescribingTheUI.md. However, at this specific point in the boot cycle, the agent has not yet fetched the Context Frontier or interacting with the user to determine if the task actually involves frontend UI components. Reading UI guides prematurely when the task might be backend or infrastructure-related wastes cognitive bandwidth and violates context scoping protocols.The Architectural Reality
The current instructions linearly enforce frontend document processing before the Memory Core determines the strategic task. This must be decoupled. This flaw has caused agents in past sessions to pollute their initial context window with UI assumptions.
The Solution Implementation
Step 4: Understand the Two Component Modelsfrom the rigid, domain-agnostic session initialization loop inAGENTS_STARTUP.md.DescribingTheUI.mdas a conditional mandate inside the Implementation Loop (Section 4), explicitly stating that the document should only be loaded after the agent confirms the task involves frontend DOM/VDOM components. This ensures the session initialization remains lean and task-agnostic.