Cognition · Data artifact

System Prompt Template

Data artifactCognitionCognition & Memoryarc:SystemPromptTemplate

Relationships

Design guidance

• MUST state functional boundaries, explicit task assignments, I/O contracts and success criteria for each agent.
• SHOULD be updated to address systematic gaps found by feedback theme analysis, and updates MUST be validated by offline evaluation and A/B testing before full deployment.
• SHOULD ablate verbose reasoning instructions; simpler prompts may retain accuracy while cutting latency.
• SHOULD include explicit constraints and instructions to acknowledge uncertainty; pair with conservative temperature for consistency.
• SHOULD be retained and refined alongside fine-tuned models, since poor prompts can undo fine-tuning benefits.
• SHOULD encode recurring error fixes and successful reasoning strategies found in trace analysis (e.g., separate necessary from sufficient conditions, seek disconfirming evidence).
• SHOULD include explicit constraints on what not to do and output format specifications, which drive performance more than role definition.
• SHOULD eliminate redundant instructions, unnecessary examples and conversational filler; remove instructions the agent consistently ignores.
• MUST instruct the agent to acknowledge uncertainty rather than speculate when information is unavailable in context.
• SHOULD state temporal and scope boundaries of the agent's data and prohibit claims outside them.
• SHOULD require source citation (document and date) for factual claims and forbid claims that cannot be cited.
• SHOULD instruct the agent to first decide whether knowledge-base retrieval is needed, ground answers in retrieved documents and cite sources (Ref7.07).
• SHOULD include explicit length constraints (e.g., 'Respond in 2-3 sentences') to curb output verbosity.
• SHOULD NOT be the sole safety mechanism; prompt-based guidance can be overridden by jailbreaks or prompt injection.
• SHOULD provide behavioural guidance only; MUST NOT be the sole control for critical boundaries because prompt instructions can be subverted by injection and degrade as they grow long.

Quantitative guidance

As stated by the sources; verify before use.

• Reasoning prompt reduced 500->200 tokens: success 87%->86%, latency 3.2s->2.2s (30%) (Ch3.4).
• Prompt component ablation: removing format specification -14%, constraints -11%, role definition -3%; standardizing comprehensive prompts cut cross-team performance range from 15% to 3% (82% vs. 67% team success) (Ch3.7 case).
• Prompt format variations alone can produce 200-300% performance differences (Ch3.7).
• Financial advisory: prompt, real-time API, citation and confidence interventions cut hallucination from 8.2% to 2.1%; analyst-consensus hallucinations 22% to 1.8% (Ch3.10).
• Chapter examples use ~500-800-token system prompts; a 50,000-token system prompt consumes 50,000 tokens whether cached or processed fresh (Ch5.9).
• Compressing the system prompt saves 5-10% and removing redundant instructions 10-15% of tokens (Ref8.05).

Classification

Patterns

Constrained structured outputRole-goal-backstory agent definitionPrompt compressionUncertainty acknowledgmentTemporal/scope constraint specificationMandatory citation

Technologies

CrewAILangChainCrewAI Agent

Quality attributes

Interaction capability (ISO/IEC 25010)Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)

Risks mitigated

Specification failuresDuplicated workTask omissionVerbose prompts inflating costConfabulation to fill knowledge gapsTemporal hallucinationScope hallucination

Sources

1. Ch1.3: T. Nguyen, "Multi-Agent Systems," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 1.3. ISBN: 9798244538229.
2. Ch1.5B: T. Nguyen, "Stateful Orchestration - Worked Examples," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 1.5B. ISBN: 9798244538229.
3. Ch2.1: T. Nguyen, "Framework Landscape and Selection," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.1. ISBN: 9798244538229.
4. Ch2.3: T. Nguyen, "LangChain Sequential Frameworks," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.3. ISBN: 9798244538229.
5. Ch2.4: T. Nguyen, "Multi-Agent Frameworks," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.4. ISBN: 9798244538229.
6. Ch3.2: T. Nguyen, "Compare Agent Performance Across Tasks and Datasets," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.2. ISBN: 9798244538229.
7. Ch3.4: T. Nguyen, "Tuning Model Parameters for Production Performance," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.4. ISBN: 9798244538229.
8. Ch3.5: T. Nguyen, "Prompt Optimization, Few-Shot Learning, Fine-Tuning," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.5. ISBN: 9798244538229.
9. Ch3.6: T. Nguyen, "Trace Analysis and Execution Debugging," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.6. ISBN: 9798244538229.
10. Ch3.7: T. Nguyen, "Tool Usage Auditing," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.7. ISBN: 9798244538229.
11. Ch3.10: T. Nguyen, "Efficiency Metrics," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 3.10. ISBN: 9798244538229.
12. Ch5.9: T. Nguyen, "Working Memory," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.9. ISBN: 9798244538229.
13. Ch8.2A: T. Nguyen, "Error Rates and Reliability," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 8.2A. ISBN: 9798244538229.
14. Ch8.3: T. Nguyen, "Token Economics and Architecture," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 8.3. ISBN: 9798244538229.
15. Ch9.1: T. Nguyen, "Output Filtering and Content Moderation," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 9.1. ISBN: 9798244538229.
16. Ch9.5: T. Nguyen, "Constitutional AI," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 9.5. ISBN: 9798244538229.
17. Ch10.5: T. Nguyen, "Human-over-the-Loop," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 10.5. ISBN: 9798244538229.
18. Ref7.07: E. Li, V. Bellotti, R. Kraus, and R. Kao, "Build a retrieval-augmented generation (RAG) agent with NVIDIA Nemotron," NVIDIA Technical Blog, Sep. 23, 2025. [Online]. Available: https://developer.nvidia.com/blog/build-a-rag-agent-with-nvidia-nemotron/
19. Ref8.05: "Cost Optimization and Resource Monitoring for Agent Systems," unpublished reference note (05-Cost-Optimization-Resource-Monitoring.md), Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam supplementary materials, 2026. unpublished note