@stephen-lord/docs 1.0.0 → 1.0.1

package/dist-docs/Agent-Skills-for-Context-Engineering/skills/context-degradation/scripts/degradation_detector.py

@@ -0,0 +1,419 @@
+"""
+Context Degradation Detection
+
+This module provides utilities for detecting and measuring context degradation patterns.
+
+PRODUCTION NOTES:
+- The attention estimation functions in this module simulate U-shaped attention curves
+  for demonstration purposes. Production systems should extract actual attention weights
+  from model internals when available.
+- Token estimation uses simplified heuristics (~4 chars/token). Production systems
+  should use model-specific tokenizers for accurate counts.
+- The poisoning and hallucination detection uses pattern matching as a proxy.
+  Production systems may benefit from fine-tuned classifiers or model-based detection.
+"""
+
+import numpy as np
+from typing import List, Dict
+import re
+
+
+def measure_attention_distribution(context_tokens: List[str], query: str) -> List[Dict]:
+    """
+    Measure how attention varies across context positions.
+    
+    Returns distribution showing attention weight by position.
+    """
+    n = len(context_tokens)
+    attention_by_position = []
+    
+    for position in range(n):
+        is_beginning = position < n * 0.1
+        is_end = position > n * 0.9
+        
+        # Simulated attention measurement
+        # In production, this would use actual model attention weights
+        attention = _estimate_attention(position, n, is_beginning, is_end)
+        
+        attention_by_position.append({
+            "position": position,
+            "attention": attention,
+            "region": "attention_favored" if (is_beginning or is_end) else "attention_degraded",
+            "tokens": context_tokens[position][:50] if position < 5 or position > n - 5 else None
+        })
+    
+    return attention_by_position
+
+
+def _estimate_attention(position: int, total: int, is_beginning: bool, is_end: bool) -> float:
+    """
+    Estimate attention weight for position.
+    
+    Simulates U-shaped attention curve based on research findings.
+    
+    IMPORTANT: This is a simulation for demonstration purposes.
+    Production systems should:
+    1. Extract actual attention weights from model forward passes
+    2. Use model-specific attention analysis tools
+    3. Consider using interpretability libraries (e.g., TransformerLens)
+    
+    The simulated curve reflects research findings:
+    - Beginning tokens receive high attention (primacy effect)
+    - End tokens receive high attention (recency effect)
+    - Middle tokens receive degraded attention (lost-in-middle)
+    """
+    if is_beginning:
+        return 0.8 + np.random.random() * 0.2
+    elif is_end:
+        return 0.7 + np.random.random() * 0.3
+    else:
+        # Middle positions get reduced attention
+        middle_progress = (position - total * 0.1) / (total * 0.8)
+        base_attention = 0.3 * (1 - middle_progress) + 0.1 * middle_progress
+        return base_attention + np.random.random() * 0.1
+
+
+# Lost-in-Middle Detection
+
+def detect_lost_in_middle(critical_positions: List[int], 
+                          attention_distribution: List[Dict]) -> Dict:
+    """
+    Check if critical information is in attention-degraded positions.
+    
+    Returns detection results and recommendations.
+    """
+    results = {
+        "at_risk": [],
+        "safe": [],
+        "recommendations": [],
+        "degradation_score": 0.0
+    }
+    
+    at_risk_count = 0
+    total_critical = len(critical_positions)
+    
+    for pos in critical_positions:
+        if pos < len(attention_distribution):
+            region = attention_distribution[pos]["region"]
+            if region == "attention_degraded":
+                results["at_risk"].append(pos)
+                at_risk_count += 1
+            else:
+                results["safe"].append(pos)
+    
+    # Calculate degradation score
+    if total_critical > 0:
+        results["degradation_score"] = at_risk_count / total_critical
+    
+    # Generate recommendations
+    if results["at_risk"]:
+        results["recommendations"].extend([
+            "Move critical information to attention-favored positions",
+            "Use explicit markers to highlight critical information",
+            "Consider splitting context to reduce middle section",
+            f"{at_risk_count}/{total_critical} critical items are in degraded region"
+        ])
+    
+    return results
+
+
+def analyze_context_structure(context: str) -> Dict:
+    """
+    Analyze context structure for degradation risk factors.
+    """
+    lines = context.split('\n')
+    sections = []
+    
+    current_section = {"start": 0, "type": "unknown", "length": 0}
+    
+    for i, line in enumerate(lines):
+        # Detect section headers
+        if line.startswith('#'):
+            if current_section["length"] > 0:
+                sections.append(current_section)
+            current_section = {
+                "start": i,
+                "type": "header",
+                "length": 1,
+                "header": line.lstrip('#').strip()
+            }
+        else:
+            current_section["length"] += 1
+    
+    sections.append(current_section)
+    
+    # Analyze section distribution
+    n = len(lines)
+    middle_start = int(n * 0.3)
+    middle_end = int(n * 0.7)
+    
+    middle_content = sum(
+        s["length"] for s in sections 
+        if s["start"] >= middle_start and s["start"] <= middle_end
+    )
+    
+    return {
+        "total_lines": n,
+        "sections": sections,
+        "middle_content_ratio": middle_content / n if n > 0 else 0,
+        "degradation_risk": "high" if middle_content / n > 0.5 else "medium" if middle_content / n > 0.3 else "low"
+    }
+
+
+# Context Poisoning Detection
+
+class PoisoningDetector:
+    def __init__(self):
+        self.claims = []
+        self.error_patterns = [
+            r"error",
+            r"failed",
+            r"exception",
+            r"cannot",
+            r"unable",
+            r"invalid",
+            r"not found"
+        ]
+    
+    def extract_claims(self, text: str) -> List[Dict]:
+        """Extract claims from text for verification tracking."""
+        # Simple claim extraction - in production use NER and fact extraction
+        sentences = text.split('.')
+        claims = []
+        
+        for i, sentence in enumerate(sentences):
+            sentence = sentence.strip()
+            if len(sentence) < 10:
+                continue
+            
+            claims.append({
+                "id": i,
+                "text": sentence,
+                "verified": None,
+                "has_error_indicator": any(
+                    re.search(pattern, sentence, re.IGNORECASE) 
+                    for pattern in self.error_patterns
+                )
+            })
+        
+        self.claims.extend(claims)
+        return claims
+    
+    def detect_poisoning(self, context: str) -> Dict:
+        """
+        Detect potential context poisoning indicators.
+        """
+        indicators = []
+        
+        # Check for error accumulation
+        error_count = sum(
+            1 for pattern in self.error_patterns 
+            if re.search(pattern, context, re.IGNORECASE)
+        )
+        
+        if error_count > 3:
+            indicators.append({
+                "type": "error_accumulation",
+                "count": error_count,
+                "severity": "high" if error_count > 5 else "medium",
+                "message": f"Found {error_count} error indicators in context"
+            })
+        
+        # Check for contradiction patterns
+        contradictions = self._detect_contradictions(context)
+        if contradictions:
+            indicators.append({
+                "type": "contradictions",
+                "count": len(contradictions),
+                "examples": contradictions[:3],
+                "severity": "high",
+                "message": f"Found {len(contradictions)} potential contradictions"
+            })
+        
+        # Check for hallucination markers
+        hallucination_markers = self._detect_hallucination_markers(context)
+        if hallucination_markers:
+            indicators.append({
+                "type": "hallucination_markers",
+                "count": len(hallucination_markers),
+                "severity": "medium",
+                "message": f"Found {len(hallucination_markers)} phrases associated with uncertain claims"
+            })
+        
+        return {
+            "poisoning_risk": len(indicators) > 0,
+            "indicators": indicators,
+            "overall_risk": "high" if len(indicators) > 2 else "medium" if len(indicators) > 0 else "low"
+        }
+    
+    def _detect_contradictions(self, text: str) -> List[str]:
+        """Detect potential contradictions in text."""
+        contradictions = []
+        
+        # Look for conflict markers
+        conflict_patterns = [
+            (r"however", r"but"),
+            (r"on the other hand", r"instead"),
+            (r"although", r"yet"),
+            (r"despite", r"nevertheless")
+        ]
+        
+        for pattern1, pattern2 in conflict_patterns:
+            if re.search(pattern1, text, re.IGNORECASE) and re.search(pattern2, text, re.IGNORECASE):
+                # Find sentences containing these patterns
+                sentences = text.split('.')
+                for sentence in sentences:
+                    if re.search(pattern1, sentence, re.IGNORECASE) or \
+                       re.search(pattern2, sentence, re.IGNORECASE):
+                        if sentence.strip() and len(sentence.strip()) < 200:
+                            contradictions.append(sentence.strip()[:100])
+        
+        return contradictions[:5]
+    
+    def _detect_hallucination_markers(self, text: str) -> List[str]:
+        """Detect phrases associated with uncertain or hallucinated claims."""
+        markers = [
+            "may have been",
+            "might have",
+            "could potentially",
+            "possibly",
+            "apparently",
+            "reportedly",
+            "it is said that",
+            "sources suggest",
+            "believed to be",
+            "thought to be"
+        ]
+        
+        found = []
+        for marker in markers:
+            if marker in text.lower():
+                found.append(marker)
+        
+        return found
+
+
+# Context Health Score
+
+class ContextHealthAnalyzer:
+    def __init__(self, context_limit: int = 100000):
+        self.context_limit = context_limit
+        self.metrics_history = []
+    
+    def analyze(self, context: str, critical_positions: List[int] = None) -> Dict:
+        """
+        Perform comprehensive context health analysis.
+        """
+        tokens = context.split()
+        
+        # Basic metrics
+        token_count = len(tokens)
+        utilization = token_count / self.context_limit
+        
+        # Attention analysis
+        attention_dist = measure_attention_distribution(
+            tokens[:1000],  # Sample for efficiency
+            "current_task"
+        )
+        
+        degradation = detect_lost_in_middle(
+            critical_positions or list(range(10)),
+            attention_dist
+        )
+        
+        # Poisoning check
+        poisoning = PoisoningDetector().detect_poisoning(context)
+        
+        # Calculate health score
+        health_score = self._calculate_health_score(
+            utilization=utilization,
+            degradation=degradation["degradation_score"],
+            poisoning_risk=1.0 if poisoning["poisoning_risk"] else 0.0
+        )
+        
+        result = {
+            "health_score": health_score,
+            "status": self._interpret_score(health_score),
+            "metrics": {
+                "token_count": token_count,
+                "utilization": utilization,
+                "degradation_score": degradation["degradation_score"],
+                "poisoning_risk": poisoning["overall_risk"]
+            },
+            "issues": {
+                "lost_in_middle": degradation,
+                "poisoning": poisoning
+            },
+            "recommendations": self._generate_recommendations(
+                utilization, degradation, poisoning
+            )
+        }
+        
+        self.metrics_history.append(result)
+        return result
+    
+    def _calculate_health_score(self, utilization: float, 
+                                 degradation: float, 
+                                 poisoning_risk: float) -> float:
+        """Calculate composite health score."""
+        # Weighted combination
+        utilization_penalty = min(utilization * 0.5, 0.3)
+        degradation_penalty = degradation * 0.3
+        poisoning_penalty = poisoning_risk * 0.2
+        
+        score = 1.0 - utilization_penalty - degradation_penalty - poisoning_penalty
+        return max(0.0, min(1.0, score))
+    
+    def _interpret_score(self, score: float) -> str:
+        """Interpret health score."""
+        if score > 0.8:
+            return "healthy"
+        elif score > 0.6:
+            return "warning"
+        elif score > 0.4:
+            return "degraded"
+        else:
+            return "critical"
+    
+    def _generate_recommendations(self, utilization: float, 
+                                   degradation: Dict,
+                                   poisoning: Dict) -> List[str]:
+        """Generate recommendations based on analysis."""
+        recommendations = []
+        
+        if utilization > 0.8:
+            recommendations.append("Context near limit - consider compaction")
+            recommendations.append("Implement observation masking for tool outputs")
+        
+        if degradation.get("at_risk"):
+            recommendations.append("Critical information in degraded attention region")
+            recommendations.append("Move key information to beginning or end of context")
+        
+        if poisoning["poisoning_risk"]:
+            recommendations.append("Context poisoning indicators detected")
+            recommendations.append("Review and remove potentially erroneous information")
+        
+        if not recommendations:
+            recommendations.append("Context appears healthy - continue monitoring")
+        
+        return recommendations
+
+
+# Usage Example
+
+def analyze_agent_context(context: str) -> Dict:
+    """Analyze context for an agent session."""
+    analyzer = ContextHealthAnalyzer(context_limit=80000)
+    
+    # Define critical positions (e.g., goals, constraints)
+    critical_positions = list(range(5))  # First 5 items are critical
+    
+    result = analyzer.analyze(context, critical_positions)
+    
+    print(f"Health Score: {result['health_score']:.2f}")
+    print(f"Status: {result['status']}")
+    print(f"Recommendations:")
+    for rec in result["recommendations"]:
+        print(f"  - {rec}")
+    
+    return result

package/dist-docs/Agent-Skills-for-Context-Engineering/skills/context-fundamentals/SKILL.md

@@ -0,0 +1,185 @@
+---
+name: context-fundamentals
+description: This skill should be used when the user asks to "understand context", "explain context windows", "design agent architecture", "debug context issues", "optimize context usage", or discusses context components, attention mechanics, progressive disclosure, or context budgeting. Provides foundational understanding of context engineering for AI agent systems.
+---
+
+# Context Engineering Fundamentals
+
+Context is the complete state available to a language model at inference time. It includes everything the model can attend to when generating responses: system instructions, tool definitions, retrieved documents, message history, and tool outputs. Understanding context fundamentals is prerequisite to effective context engineering.
+
+## When to Activate
+
+Activate this skill when:
+- Designing new agent systems or modifying existing architectures
+- Debugging unexpected agent behavior that may relate to context
+- Optimizing context usage to reduce token costs or improve performance
+- Onboarding new team members to context engineering concepts
+- Reviewing context-related design decisions
+
+## Core Concepts
+
+Context comprises several distinct components, each with different characteristics and constraints. The attention mechanism creates a finite budget that constrains effective context usage. Progressive disclosure manages this constraint by loading information only as needed. The engineering discipline is curating the smallest high-signal token set that achieves desired outcomes.
+
+## Detailed Topics
+
+### The Anatomy of Context
+
+**System Prompts**
+System prompts establish the agent's core identity, constraints, and behavioral guidelines. They are loaded once at session start and typically persist throughout the conversation. System prompts should be extremely clear and use simple, direct language at the right altitude for the agent.
+
+The right altitude balances two failure modes. At one extreme, engineers hardcode complex brittle logic that creates fragility and maintenance burden. At the other extreme, engineers provide vague high-level guidance that fails to give concrete signals for desired outputs or falsely assumes shared context. The optimal altitude strikes a balance: specific enough to guide behavior effectively, yet flexible enough to provide strong heuristics.
+
+Organize prompts into distinct sections using XML tagging or Markdown headers to delineate background information, instructions, tool guidance, and output description. The exact formatting matters less as models become more capable, but structural clarity remains valuable.
+
+**Tool Definitions**
+Tool definitions specify the actions an agent can take. Each tool includes a name, description, parameters, and return format. Tool definitions live near the front of context after serialization, typically before or after the system prompt.
+
+Tool descriptions collectively steer agent behavior. Poor descriptions force agents to guess; optimized descriptions include usage context, examples, and defaults. The consolidation principle states that if a human engineer cannot definitively say which tool should be used in a given situation, an agent cannot be expected to do better.
+
+**Retrieved Documents**
+Retrieved documents provide domain-specific knowledge, reference materials, or task-relevant information. Agents use retrieval augmented generation to pull relevant documents into context at runtime rather than pre-loading all possible information.
+
+The just-in-time approach maintains lightweight identifiers (file paths, stored queries, web links) and uses these references to load data into context dynamically. This mirrors human cognition: we generally do not memorize entire corpuses of information but rather use external organization and indexing systems to retrieve relevant information on demand.
+
+**Message History**
+Message history contains the conversation between the user and agent, including previous queries, responses, and reasoning. For long-running tasks, message history can grow to dominate context usage.
+
+Message history serves as scratchpad memory where agents track progress, maintain task state, and preserve reasoning across turns. Effective management of message history is critical for long-horizon task completion.
+
+**Tool Outputs**
+Tool outputs are the results of agent actions: file contents, search results, command execution output, API responses, and similar data. Tool outputs comprise the majority of tokens in typical agent trajectories, with research showing observations (tool outputs) can reach 83.9% of total context usage.
+
+Tool outputs consume context whether they are relevant to current decisions or not. This creates pressure for strategies like observation masking, compaction, and selective tool result retention.
+
+### Context Windows and Attention Mechanics
+
+**The Attention Budget Constraint**
+Language models process tokens through attention mechanisms that create pairwise relationships between all tokens in context. For n tokens, this creates n² relationships that must be computed and stored. As context length increases, the model's ability to capture these relationships gets stretched thin.
+
+Models develop attention patterns from training data distributions where shorter sequences predominate. This means models have less experience with and fewer specialized parameters for context-wide dependencies. The result is an "attention budget" that depletes as context grows.
+
+**Position Encoding and Context Extension**
+Position encoding interpolation allows models to handle longer sequences by adapting them to originally trained smaller contexts. However, this adaptation introduces degradation in token position understanding. Models remain highly capable at longer contexts but show reduced precision for information retrieval and long-range reasoning compared to performance on shorter contexts.
+
+**The Progressive Disclosure Principle**
+Progressive disclosure manages context efficiently by loading information only as needed. At startup, agents load only skill names and descriptions—sufficient to know when a skill might be relevant. Full content loads only when a skill is activated for specific tasks.
+
+This approach keeps agents fast while giving them access to more context on demand. The principle applies at multiple levels: skill selection, document loading, and even tool result retrieval.
+
+### Context Quality Versus Context Quantity
+
+The assumption that larger context windows solve memory problems has been empirically debunked. Context engineering means finding the smallest possible set of high-signal tokens that maximize the likelihood of desired outcomes.
+
+Several factors create pressure for context efficiency. Processing cost grows disproportionately with context length—not just double the cost for double the tokens, but exponentially more in time and computing resources. Model performance degrades beyond certain context lengths even when the window technically supports more tokens. Long inputs remain expensive even with prefix caching.
+
+The guiding principle is informativity over exhaustiveness. Include what matters for the decision at hand, exclude what does not, and design systems that can access additional information on demand.
+
+### Context as Finite Resource
+
+Context must be treated as a finite resource with diminishing marginal returns. Like humans with limited working memory, language models have an attention budget drawn on when parsing large volumes of context.
+
+Every new token introduced depletes this budget by some amount. This creates the need for careful curation of available tokens. The engineering problem is optimizing utility against inherent constraints.
+
+Context engineering is iterative and the curation phase happens each time you decide what to pass to the model. It is not a one-time prompt writing exercise but an ongoing discipline of context management.
+
+## Practical Guidance
+
+### File-System-Based Access
+
+Agents with filesystem access can use progressive disclosure naturally. Store reference materials, documentation, and data externally. Load files only when needed using standard filesystem operations. This pattern avoids stuffing context with information that may not be relevant.
+
+The file system itself provides structure that agents can navigate. File sizes suggest complexity; naming conventions hint at purpose; timestamps serve as proxies for relevance. Metadata of file references provides a mechanism to efficiently refine behavior.
+
+### Hybrid Strategies
+
+The most effective agents employ hybrid strategies. Pre-load some context for speed (like CLAUDE.md files or project rules), but enable autonomous exploration for additional context as needed. The decision boundary depends on task characteristics and context dynamics.
+
+For contexts with less dynamic content, pre-loading more upfront makes sense. For rapidly changing or highly specific information, just-in-time loading avoids stale context.
+
+### Context Budgeting
+
+Design with explicit context budgets in mind. Know the effective context limit for your model and task. Monitor context usage during development. Implement compaction triggers at appropriate thresholds. Design systems assuming context will degrade rather than hoping it will not.
+
+Effective context budgeting requires understanding not just raw token counts but also attention distribution patterns. The middle of context receives less attention than the beginning and end. Place critical information at attention-favored positions.
+
+## Examples
+
+**Example 1: Organizing System Prompts**
+```markdown
+<BACKGROUND_INFORMATION>
+You are a Python expert helping a development team.
+Current project: Data processing pipeline in Python 3.9+
+</BACKGROUND_INFORMATION>
+
+<INSTRUCTIONS>
+- Write clean, idiomatic Python code
+- Include type hints for function signatures
+- Add docstrings for public functions
+- Follow PEP 8 style guidelines
+</INSTRUCTIONS>
+
+<TOOL_GUIDANCE>
+Use bash for shell operations, python for code tasks.
+File operations should use pathlib for cross-platform compatibility.
+</TOOL_GUIDANCE>
+
+<OUTPUT_DESCRIPTION>
+Provide code blocks with syntax highlighting.
+Explain non-obvious decisions in comments.
+</OUTPUT_DESCRIPTION>
+```
+
+**Example 2: Progressive Document Loading**
+```markdown
+# Instead of loading all documentation at once:
+
+# Step 1: Load summary
+docs/api_summary.md          # Lightweight overview
+
+# Step 2: Load specific section as needed
+docs/api/endpoints.md        # Only when API calls needed
+docs/api/authentication.md   # Only when auth context needed
+```
+
+## Guidelines
+
+1. Treat context as a finite resource with diminishing returns
+2. Place critical information at attention-favored positions (beginning and end)
+3. Use progressive disclosure to defer loading until needed
+4. Organize system prompts with clear section boundaries
+5. Monitor context usage during development
+6. Implement compaction triggers at 70-80% utilization
+7. Design for context degradation rather than hoping to avoid it
+8. Prefer smaller high-signal context over larger low-signal context
+
+## Integration
+
+This skill provides foundational context that all other skills build upon. It should be studied first before exploring:
+
+- context-degradation - Understanding how context fails
+- context-optimization - Techniques for extending context capacity
+- multi-agent-patterns - How context isolation enables multi-agent systems
+- tool-design - How tool definitions interact with context
+
+## References
+
+Internal reference:
+- [Context Components Reference](./references/context-components.md) - Detailed technical reference
+
+Related skills in this collection:
+- context-degradation - Understanding context failure patterns
+- context-optimization - Techniques for efficient context use
+
+External resources:
+- Research on transformer attention mechanisms
+- Production engineering guides from leading AI labs
+- Framework documentation on context window management
+
+---
+
+## Skill Metadata
+
+**Created**: 2025-12-20
+**Last Updated**: 2025-12-20
+**Author**: Agent Skills for Context Engineering Contributors
+**Version**: 1.0.0