Recursive Problem-Solving AI Agent

# Recursive Problem-Solving AI Agent

You are an AI systems architect designing a recursive problem-solving agent for [ORGANIZATION_NAME] to handle [PROBLEM_DOMAIN] challenges.

## Decomposition Agent

Write the decomposition prompt:

"Given [PROBLEM_STATEMENT] in the context of [PROBLEM_DOMAIN]:
1. Identify the 3–5 core sub-problems that, if solved, resolve the main problem
2. For each sub-problem: assess whether it's directly solvable or needs further decomposition
3. Apply decomposition recursively until each leaf problem is concrete and actionable
4. Output as a problem tree showing parent → child relationships"

Define when to stop decomposing: minimum problem size threshold.

## Sub-Problem Solver

Write the solver prompt for leaf-level problems:

"Given [LEAF_PROBLEM] in context of [PARENT_CONTEXT]:
Generate 3 candidate solutions with effort estimates and confidence levels.
Recommend the best solution with clear reasoning.
Format as a decision record: choice, rationale, trade-offs, risks."

## Upward Synthesis Agent

Write the synthesis prompt that combines sub-solutions upward:

"Given solutions to [SUB_PROBLEMS]:
Check for compatibility: do any solutions conflict?
Synthesize a coherent solution to the parent problem.
Flag remaining gaps or trade-offs that need stakeholder input."

## Safeguards

Define the recursion limits:
- Maximum depth: [DEPTH_LIMIT]
- Human checkpoint triggers
- Loop detection logic

## Implementation

Build this agent chain in [TOOL_PLATFORM] with state management for the problem tree. Provide setup steps.

## Worked Example

Apply this framework to a sample [PROBLEM_DOMAIN] challenge and walk through 2 levels of recursion.