prompt architecture

18 articles · 15 co-occurring · 0 contradictions · 0 briefs

Describes sophisticated multi-level prompt structure: startup-injected skills (preloaded context) vs task-triggered contexts (context: fork).

Related concepts

multi agent orchestration 10 tool integration patterns 9 context window management 9 context window optimization 7 state management 4 model selection strategy 3 prompt engineering 2 multi turn conversation management 2 agent autonomy 2 workflow automation 1 tool use context management 1 tool interface design 1 testing strategies 1 task decomposition 1 system prompt design 1

Evidence chain (18 articles, showing 18)

@dani_avila7: If you're defining a role: subagent with skills. The agent is the long-lived ... extends

Describes sophisticated multi-level prompt structure: startup-injected skills (preloaded context) vs task-triggered contexts (context: fork).

@IntuitMachine: # I. A Text That Reads Itself Into Being extends

Moves beyond technical structure to information sequencing and narrative framing as architectural choices

Context Engineering Guide 2026: GPT-5, Claude 4.6, Gemini Prompts That Work extends

Proposes 6-component universal template for 2026, positioning this as evolution beyond traditional prompt engineering. Suggests architecture as primary concern.

@NirDiamantAI: Claude Advisor pairs Opus as strategist with Sonnet or Haiku as executor. extends

The advisor-executor split requires different prompt structures for each tier: strategic planning prompt for Opus, execution prompts for Sonnet/Haiku

The hottest discussion in AI is about "context engineering". | Philipp ... extends

Context engineering goes beyond single-prompt design to systematic context structuring across agent lifecycle.

@IntuitMachine: I have reached the conclusion that frontier AI models are incapable of dealin... supports

QPT as a framework is fundamentally about prompt architecture—structuring the problem representation so the model can reason about its own reasoning

Architecting Smarter Multi-Agent Systems with Context Engineering example_of

Role definitions (Planner, Researcher, etc.) are a prompt architecture pattern—each role needs distinct system instructions reflecting its context access

@bcherny: 👋 Appreciate the feedback. extends

The four-point solution approach (autonomy, generalization, opt-in, monitoring) is fundamentally about system-level prompt architecture—how to encode context about capabilities into model behavior

@shao__meng: Anthropic 增长团队 @helloitsaustin 提出了一套超越"用 AI 加速现有工作"的系统性框架，一共四个维度按价值层级递进。 example_of

The shift from 'generate variant faster' to 'generate 20 angles simultaneously' is a prompt architecture decision—how you frame the task determines output quality and value.

@shao__meng: 用单线程追问根治"需求理解偏差" extends

Single-threaded questioning discipline functions as a prompt decomposition strategy—preventing wide scatter and maintaining depth of context

Stop using Claude as just a chatbot—MCP changes everything supports

With MCP connectors available, system prompts can reference specific tools/capabilities with confidence they'll work predictably. Protocol standardization enables more reliable prompt design.

Multi AI Agent Systems with crewAI (Short Course) | Coursera extends

Role-based agents require role definition through prompting, but course doesn't reveal how roles are specified or coordinated

@cristofrcharles: If you have been using Claude Code CLI like I have and run into constant poor... supports

Different tool interfaces likely wrap the same model with different system prompts and task structuring, demonstrating that prompt architecture (not just capability) drives output quality.

@dani_avila7: One important point Boris makes in this post: generalize features instead of ... example_of

CLAUDE.md mentioned as mechanism for enforcing standards across team—this is prompt/context architecture pattern

@FarzaTV: This is Farzapedia. extends

System prompt likely describes wiki structure and navigation. But broader insight is that context structure itself is a form of implicit prompt engineering.

@badlogicgames: Call out now also part of pi's docs. Just do it. supports

Understanding how agents actually respond to prompts requires seeing full traces. Prompt design improves when informed by real execution data.

2025 Year in Review | Victor Dibia extends

Multi-agent systems require explicit prompt design for each agent role; 15-chapter book likely covers agent persona/instruction patterns

AI Agent Protocol Ecosystem Map 2026: Complete Visual extends

Protocol-level architecture decisions (which protocols to use) cascade down to prompt architecture and system message design. Article doesn't discuss prompts but protocol choices constrain prompt stra

query this concept

$ db.articles("prompt-architecture")

$ db.cooccurrence("prompt-architecture")

$ db.contradictions("prompt-architecture")