Should the Agent Loop Continue

easyPROAgent Loop Control & PlanningAI AgentsAI Agents 1

Problem Statement

Decide whether a ReAct-style agent loop should take another step, given step and budget guards.

Background

The loop continues only if all hold: the goal is not yet done, the step count is below max_steps, and tokens used are below max_tokens. Any single violation stops the loop.

Your Task

def should_continue(done, steps, tokens, max_steps, max_tokens):

Return True if the loop should continue, else False.

Input Format

done (bool), steps (int), tokens (int), max_steps (int), max_tokens (int).

Output Format

A boolean.

Sample

print(should_continue(False, 3, 100, 10, 1000))

Output:

True

Example

Input

print(should_continue(False, 3, 100, 10, 1000))

Output

True

Reasoning

Inputs: done says whether the agent has already reached its goal; steps is how many think-act steps it has taken so far and max_steps is the step budget; tokens is how many tokens it has spent so far and max_tokens is the token budget.
Goal guard: done is False, so the task is unfinished and there is still a reason to keep going.
Step guard: $3 < 10$ holds, so the agent has used 3 of its 10 steps and has 7 left.
Token guard: $100 < 1000$ holds, so only 10% of the token budget is spent.
Output: all three guards pass at once, so the answer is True and the agent takes another step. If any one guard fails, the answer is False.
Notice: the budget comparisons are strict. With $3$ replaced by $10$ , the agent has used exactly its step budget, $10 < 10$ is false, and the loop stops even though the goal is unfinished and plenty of tokens remain.

Constraints

Continue only if not done and steps < max_steps and tokens < max_tokens.
Any failed condition returns False.

Learn this in

The Agent Loop & ReAct FoundationsAgent Engineering Study Plan Planning Patterns: Plan-and-Execute, ReWOO, HTNAgent Engineering Study Plan Workflow Patterns & LangGraph Stateful GraphsAgent Engineering Study Plan Multi-Agent Frameworks: AutoGen, CrewAI, OpenAI Agents SDK, Claude Agent SDKAgent Engineering Study Plan

More from AI Agents 1

Cheapest Cache TTL Meeting a Freshness SLAhard Bootstrap Confidence Interval Width for Eval Accuracymedium Route to the Cheapest Model Meeting Qualitymedium pass@k Estimatormedium Savings from a Semantic Cachemedium

Concept primerPREVIEW

1. Background Knowledge

In ReAct (Reasoning + Acting) agent architectures, an LLM alternates between reasoning steps and tool-calling actions. Because each iteration consumes computational resources and API tokens, production systems wrap the loop in guard conditions that enforce hard limits. These guards prevent runaway loops, cost overruns, and infinite recursion when the model fails to converge on a solution.

The three primary guards are:

Completion flag: A boolean indicating whether the agent has determined the goal is achieved.
Step budget: A maximum number of reasoning/action iterations allowed.
Token budget: A maximum number of tokens (input + output) consumed across all iterations.

The loop continues only when all guards are satisfied simultaneously. This is a logical AND of multiple independent constraints. If any single guard is violated, the loop must halt immediately. This pattern is common in any iterative agent framework, including AutoGPT, LangChain agents, and custom ReAct implementations.

2. Algorithm Approach

This is a straightforward boolean guard evaluation problem. The approach is:

Evaluate each guard condition independently.
Combine them with logical AND (and in Python).
Return the combined result.

There is no iteration, recursion, or data structure manipulation. The entire logic reduces to a single boolean expression. The key insight is recognizing that "the loop continues only if all hold" maps directly to a conjunction of conditions.

3. Step-by-Step Strategy

Identify the completion check: The loop should stop if done is True. So the continuation condition for this guard is not done.
Identify the step budget check: The loop should stop if steps has reached or exceeded max_steps. So the continuation condition is steps < max_steps.
Identify the token budget check: The loop should stop if tokens has reached or exceeded max_tokens. So the continuation condition is tokens < max_tokens.
Combine with AND: The loop continues only if all three conditions are true. In Python, this is a single and chain.
Return the boolean result: No additional logic is needed.

PRO

Continue the full explanation

You're reading the free preview. Unlock the complete walkthrough, the code editor, test runner and reference solution with Premium.

PRO

Editor locked

The code editor is locked for Pro problems. It is only available for free problems. Please upgrade to gain access to the code editor for all problems.

solution.py

Python 3.13

Test results