Learning Context Engineering with AWS Kiro

Kiro Meetup Japan #1 (https://findy.connpass.com/event/365956/)
Findy × AI駆動開発勉強会

Japanese ver. : https://speakerdeck.com/oikon48/kirotoxue-bukontekisutoenziniaringu

X(Twitter)：https://x.com/oikon48

Transcript

1. Learning Context Engineering with Kiro Kiro Meetup Japan #1 2025/09/03 Speaker: Oikon
2. Who are you? Oikon X: @oikon48 Software Engineer R&D at IT Company Hobbies: Personal Development, AI Tool Research Kiro Steering, Kiro and AI Tools, Trends in AI Coding Tools 1
3. What We’ll Talk Today 1. About Context Engineering 2. Practical Context Engineering 3. Context Engineering Implementation with Kiro Today, I'll present using Kiro's Spec-Driven Development (SDD) style LT 2
4. I Want to Share About Context Engineering and Kiro 3
5. Spec-Driven Development (SDD) Process requirements.md Requirements Definition design.md Technical Design tasks.md Task Breakdown 4
9. 1. Understanding Context Engineering 8
10. 1.1 Why is Context Engineering Necessary? Prompt -> Context Prompt engineering was a technique for obtaining simple outputs from simple inputs To achieve more complex outputs, we need context management techniques for more complex inputs Context Engineering: Bringing Engineering to Prompts 9
11. 1.2 Context Window Size and Contents Context Window Current major AI models have context windows of 200k to 1M tokens Components: ・System Prompt ・Memory documents ・Tool information (e.g. MCP) ・User Prompt (Prompt Engineering) ・Conversation (LLM input, output) .etc Context Engineering for Agents 10
12. 1.3 Context Engineering Challenges Traditional Approaches ・Improvement of model performance ・Expanded context windows Challenges ・Context Distraction ・Multi-turn conversation Issue ・Lost in the middle ・Context Window Overflow .etc… Even with expanded context windows, not all problems are solved. Instead of waiting for AI evolution, we need to think context engineering Cline v3.25 Context Engineering: Bringing Engineering to Prompts 11
13. 2. Practical Context Engineering 12
14. 2.1 High-Quality Context Creation Techniques Representative Approaches Planning Setup (ToDo List) Clear prioritization Task granularity adjustment Provide only relevant information Information segmentation Limit information sources Discard/compress unnecessary context Periodic reminders of important information External memory file creation Create guardrails outside context Automated execution scripts .etc… Too much to do!!! Context Engineering for Agents Context Engineering for AI Agents: Lessons from Building Manus 13
15. Context Engineering is Challenging… Many context engineering approaches exist However, it’s difficult for individuals and teams to set up development environments while being conscious of these approaches🤔 Kiro supports context engineering 14
16. 3. Context Engineering Implementation with Kiro 15
17. 3.1 Task Breakdown with SDD 【Features】 Document creation for each specification Requirements definition (requirements.md) Design specification (design.md) Task list (tasks.md) AI agents provide planning design templates and execution support 【Context Engineering Benefits】 Reduce irrelevant information to counter Context Distraction Appropriate execution granularity to prevent Lost in the middle and Multi-turn issues 16
18. 3.2 Context Assistance with Hooks 【Features】 Automatic execution of natural language instructions on trigger events File creation File saving File deletion 【Context Engineering Benefits】 Effective context standardization Independent context execution reduces context window pressure 17
19. 3.3 Persistent Context Management with Steering 【Features】 Divided persistent context for projects Product information (product.md) Project structure (structure.md) Technology stack (tech.md) Limited context usage through File Matching 【Context Engineering Benefits】 Divided context management and flexible adaptation Reduce irrelevant information to counter Context Distraction Kiro Steering 18
20. X Post 19
21. 4. Summary Kiro excels at improving input quality Kiro provides insights into context engineering We still need to engage with context engineering X: @oikon48 20
22. References Kiro Context Engineering: Bringing Engineering to Prompts Context Engineering for Agents Context Engineering 12 Factor Agents: Own Your Context Window Cline v3.25 Kiro in Context Engineering Context Engineering for AI Agents: Lessons from Building Manus Claude Code Spec Oikon’s documents: Kiro Steering Kiro and AI Tools Trends in AI Coding Tools X Post 21