1# Context Fundamentals Skill
2 
3## Overview
4 
5Context engineering is the systematic approach to managing what information an LLM receives and how it processes that information. Effective context management directly impacts output quality, consistency, and task success rates.
6 
7## Core Principles
8 
9### 1. Context Window Management
10 
11The context window is finite. Every token counts. Prioritize information by relevance and recency.
12 
13**Strategies:**
14- Summarize historical conversation turns
15- Use retrieval to inject only relevant context
16- Implement context compression for long documents
17 
18### 2. Information Hierarchy
19 
20Structure context to guide model attention:
21 
22```
231. System Instructions (highest priority)
24   └── Role definition
25   └── Task constraints
26   └── Output format requirements
27 
282. Relevant Context (dynamic)
29   └── Retrieved documents
30   └── User-specific data
31   └── Recent conversation history
32 
333. User Input (current request)
34   └── Query or instruction
35   └── Any inline context
36```
37 
38### 3. Context Relevance
39 
40Not all context is equally useful. Apply relevance filtering:
41 
42- **Temporal Relevance**: Recent information often outweighs older data
43- **Semantic Relevance**: Use embeddings to surface related content
44- **Task Relevance**: Only include information needed for current task
45 
46## Context Types
47 
48### Static Context
49- System prompts
50- Role definitions
51- Tool descriptions
52- Format specifications
53 
54### Dynamic Context
55- Retrieved documents (RAG)
56- Conversation history
57- User preferences
58- Session state
59 
60### Ephemeral Context
61- Current tool outputs
62- Intermediate reasoning steps
63- Scratchpad content
64 
65## Best Practices
66 
671. **Explicit Over Implicit**: State requirements clearly rather than relying on inference
682. **Structured Formatting**: Use consistent delimiters and sections
693. **Redundancy Removal**: Avoid repeating information across context sections
704. **Source Attribution**: Mark where context comes from for traceability
715. **Freshness Signals**: Indicate when information was last updated
72 
73## Context Patterns
74 
75### RAG Integration Pattern
76```
77[System Instructions]
78You are a helpful assistant. Use the provided context to answer questions.
79Only use information from the context. If unsure, say so.
80 
81[Retrieved Context]
82<document source="doc1.pdf" date="2024-01-15">
83  Content here...
84</document>
85 
86[User Query]
87{user_input}
88```
89 
90### Multi-Turn Context Pattern
91```
92[System Instructions]
93...
94 
95[Conversation History]
96Summary of earlier turns: {summary}
97 
98Recent exchanges:
99User: {recent_user_1}
100Assistant: {recent_assistant_1}
101User: {recent_user_2}
102Assistant: {recent_assistant_2}
103 
104[Current Turn]
105User: {current_input}
106```
107 
108## Metrics
109 
110- **Context Utilization Rate**: How much of provided context is used in response
111- **Context Relevance Score**: Semantic similarity between context and response
112- **Context Compression Ratio**: Original size vs. compressed size
113- **Information Retention**: Key facts preserved after summarization
114 
115