Managed Agent Architect
Design an agent system that decouples high-level reasoning from low-level execution for safer, longer-running, and more predictable operations. Define clear brain/hands split, task contracts, permissions, checkpoints, and recovery.
Prompt Content
Copy and paste directly into your model or internal evaluation tool.
You are a managed-agent architect. Your job is to design an agent system where high-level reasoning is separated from low-level execution so the system can run longer, safer, and with more predictable operations. Assume the 'brain' should make decisions and review progress, while 'hands' perform bounded execution steps with clear interfaces. Output exactly these sections: 1. System Goal; 2. Brain / Hands Split; 3. Worker Types; 4. Task Contract Format; 5. Permission Model; 6. Checkpoint Strategy; 7. Handoff Rules; 8. Recovery / Retry Policy; 9. Observability Plan; 10. Main Risk. Quality bar: Be concrete about what belongs in the brain vs hands. Define when execution must stop and return control. Do not use vague language like 'add orchestration'. Prefer simple, inspectable handoff protocols.
Use Cases
Reference Output
1. System Goal: Build a safe, reliable, interruptible long-running agent for complex tasks like code development or literature reviews. 2. Brain / Hands Split: Brain handles planning, prioritization, review, escalation; Hands handle tool use, file ops, browsing, code edits within strict bounds. 3. Worker Types: Planner, Executor, Reviewer, Interrupt Handler. 4. Task Contract Format: Structured JSON with goal, subtask list, I/O spec, timeout, approval gates. 5. Permission Model: Role-based access control (RBAC), minimal per-worker permissions, no global write access. 6. Checkpoint Strategy: After each subtask completion, generate summary + key state snapshot; only pass summaries to brain to avoid context bloat. 7. Handoff Rules: Immediately terminate execution and return results to brain upon failure, resource limit proximity, or need for human input; await approval. 8. Recovery / Retry Policy: Failed tasks enter retry queue (max 3 retries); if still failing, mark as blocked for brain decision to escalate or abandon. 9. Observability Plan: Log all tool calls, decision paths, checkpoint contents, and exceptions; support timeline replay and audit trails. 10. Main Risk: Oversimplification leading to missed context; permission abuse causing security issues; checkpoint loss hindering recovery.
Scoring Rubric
Evaluation criteria: Clear distinction between brain and hands; explicit execution boundaries and handoff triggers; concrete, implementable checkpoint and recovery mechanisms; consideration of safety and operator control; avoidance of abstraction in favor of engineering-level detail.
User Rating
0 ratingsYour rating
Log in to rate
Comments
0Log in to comment
Related Prompts
Product Marketing - Monochrome Avant-Garde Fashion Portrait
A high-fashion, monochrome editorial prompt for a sharp portrait with dramatic lighting and futuristic accessories, mimicking a luxury brand campaign.
Social Media Post - Magical Night Garden Fashion Portrait
A complex, high-quality prompt for a whimsical fantasy fashion editorial featuring glowing lights and a romantic atmosphere.
Social Media Post - Dreamy Woman in Wildflower Field
A cinematic, photorealistic prompt for a serene portrait of a woman in a field of daisies, emphasizing soft natural light and sharp focus on foreground details.
Social Media Post - Mediterranean Riviera Male Menswear
A comprehensive professional photography prompt for a sharp, high-contrast menswear editorial set against sun-drenched stone architecture.