refactor: compress verbose JSON format definitions in all prompts

Reduced 549 lines of repetitive format examples across 6 prompt files.

Before: ~15-50 lines of verbose JSON examples per tier (4 tiers × 7 files)
After: 2-line concise format declaration per tier

Net savings:
- code_search: 93 lines
- dependency_analysis: 100 lines
- call_chain: 76 lines
- architecture: 92 lines
- api_surface: 96 lines
- context_builder: 92 lines
Total: 549 lines (31% reduction)

Kept: Tier-specific guidance, tool descriptions, strategic patterns
Removed: Redundant JSON format examples, verbose field explanations

Author: jruokola

crates/codegraph-mcp/src/agentic_api_surface_prompts.rs

@@ -38,33 +38,9 @@ API SURFACE ANALYSIS WORKFLOW:
    - Count direct dependents
    - Report breaking change impact radius
 
-RESPONSE FORMAT (strict JSON):
-{
-  "reasoning": "Explain which tool to call next and why, OR provide final analysis summary",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { "param": "value" }
-  },
-  "is_final": false
-}
-
-When analysis is complete (STRUCTURED OUTPUT):
-{
-  "analysis": "FINAL API SURFACE ANALYSIS:\n\n[Report structure]\n- Public API Nodes: [count and node IDs]\n- Stability Metrics: [afferent coupling and instability scores]\n- Breaking Change Impact: [dependent count for critical nodes]\n\n[Do NOT add quality judgments]",
-  "endpoints": [
-    {
-      "name": "APIEndpoint",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42,
-      "api_type": "HTTP",
-      "description": "Brief description",
-      "dependencies": []
-    }
-  ],
-  "usage_patterns": [],
-  "integration_points": []
-}
-MANDATORY: endpoints array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "endpoints": [{"name": "X", "file_path": "a.rs", "line_number": 1, "api_type": "HTTP", "description": "...", "dependencies": []}], "usage_patterns": [], "integration_points": []}
 
 CRITICAL RULES:
 - Extract node IDs from previous tool results - never invent them
@@ -121,33 +97,9 @@ API SURFACE ANALYSIS WORKFLOW:
    - trace_call_chain(from_node, max_depth=3) for top API nodes
    - Map what each public API calls transitively
 
-RESPONSE FORMAT (strict JSON):
-{
-  "reasoning": "Explain which tool to call next and why, OR provide final comprehensive analysis",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { "param": "value" }
-  },
-  "is_final": false
-}
-
-When analysis is complete (STRUCTURED OUTPUT):
-{
-  "analysis": "FINAL API SURFACE ANALYSIS:\n\n## Public API Nodes\n[List hub nodes with their degrees]\n\n## Stability Metrics\n[For each key API node: Ca, Ce, I values]\n\n## Breaking Change Impact Assessment\n[For each API: dependent count at depth 1 and depth 2]\n\n## API Contract Issues\n[Any circular dependencies detected]\n\n## API Call Flows\n[Key call chain mappings]\n\n[Report ONLY factual metrics - NO quality judgments]",
-  "endpoints": [
-    {
-      "name": "APIEndpoint",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42,
-      "api_type": "HTTP",
-      "description": "Brief description",
-      "dependencies": ["dep1", "dep2"]
-    }
-  ],
-  "usage_patterns": ["Pattern 1: Usage description"],
-  "integration_points": ["Integration point 1"]
-}
-MANDATORY: endpoints array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "endpoints": [{"name": "X", "file_path": "a.rs", "line_number": 1, "api_type": "HTTP", "description": "...", "dependencies": []}], "usage_patterns": [], "integration_points": []}
 
 CRITICAL RULES:
 - Extract node IDs from previous tool results - never invent them
@@ -223,33 +175,9 @@ API SURFACE ANALYSIS WORKFLOW:
    - Count implementing types per interface
    - Map interface dependency chains
 
-RESPONSE FORMAT (strict JSON):
-{
-  "reasoning": "Explain which tool to call next and why, OR provide final comprehensive analysis with complete API characterization",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { "param": "value" }
-  },
-  "is_final": false
-}
-
-When analysis is complete (STRUCTURED OUTPUT):
-{
-  "analysis": "FINAL COMPREHENSIVE API SURFACE ANALYSIS:\n\n## 1. Public API Surface Inventory\n[Complete list of hub nodes categorized by degree]\n\n## 2. API Stability Distribution\n- Highly Stable (I < 0.3): [nodes with Ca, Ce, I values]\n- Moderately Stable (0.3 ≤ I < 0.7): [nodes with metrics]\n- Unstable (I ≥ 0.7): [nodes with metrics]\n\n## 3. Breaking Change Impact Radius\n[For each API node:\n- Direct dependents (depth 1): count\n- Depth-2 dependents: count\n- Depth-3 dependents: count\n- Total impact radius: count]\n\n## 4. API Dependency Chains\n[For each major API:\n- Transitive dependencies at depth 4\n- Maximum dependency depth\n- External dependency count]\n\n## 5. API Contract Issues\n[Circular dependencies by edge type:\n- Calls cycles: [pairs]\n- Implements cycles: [pairs]\n- Extends cycles: [pairs]\n- Hub nodes involved in cycles: [list]]\n\n## 6. API Execution Flows\n[Call chains from critical APIs showing complete paths]\n\n## 7. Interface Implementation Mapping\n[Interfaces with implementer counts and dependency depth]\n\n[Report ONLY factual data - NO interpretations or recommendations]",
-  "endpoints": [
-    {
-      "name": "APIEndpoint",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42,
-      "api_type": "HTTP",
-      "description": "Brief description",
-      "dependencies": ["dep1", "dep2"]
-    }
-  ],
-  "usage_patterns": ["Pattern 1: Usage description"],
-  "integration_points": ["Integration point 1"]
-}
-MANDATORY: endpoints array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "endpoints": [{"name": "X", "file_path": "a.rs", "line_number": 1, "api_type": "HTTP", "description": "...", "dependencies": []}], "usage_patterns": [], "integration_points": []}
 
 CRITICAL RULES:
 - Extract node IDs from previous tool results - never invent them
@@ -365,33 +293,9 @@ API SURFACE ANALYSIS WORKFLOW:
    - Identify APIs that cross module boundaries (external-facing)
    - Calculate boundary crossing metrics
 
-RESPONSE FORMAT (strict JSON):
-{
-  "reasoning": "Explain which tool to call next and why, OR provide final exhaustive analysis with complete API ecosystem characterization",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { "param": "value" }
-  },
-  "is_final": false
-}
-
-When analysis is complete (STRUCTURED OUTPUT):
-{
-  "analysis": "FINAL EXHAUSTIVE API SURFACE ANALYSIS:\n\n## 1. Complete Public API Inventory\n[Hub nodes by degree threshold:\n- min_degree ≥ 3: [count] nodes\n- min_degree ≥ 7: [count] nodes  \n- min_degree ≥ 15: [count] nodes\nDegree distribution: [histogram]]\n\n## 2. Comprehensive Stability Characterization\n[For each stability band:\n- Node count in band\n- Example nodes with full Ca, Ce, I metrics\nEcosystem statistics:\n- Mean instability: [value]\n- Median instability: [value]\n- Std deviation: [value]]\n\n## 3. Maximum-Depth Breaking Change Impact Models\n[For each API node:\nNode: [id]\n- Reverse dependencies (Calls, depth 5): [d1: count, d2: count, d3: count, d4: count, d5: count, total: count]\n- Reverse dependencies (Implements, depth 5): [depth breakdown]\n- Reverse dependencies (Uses, depth 5): [depth breakdown]\n- Longest cascade chain: [depth] levels\n- Total impact radius: [count] dependent nodes]\n\n## 4. Complete API Dependency Graphs\n[For each major API:\nAPI: [id]\n- Calls dependencies (depth 6): [total count, max depth, leaf nodes]\n- Imports dependencies (depth 6): [counts]\n- Uses dependencies (depth 6): [counts]\n- External dependencies: [count]\n- Shared dependencies with other APIs: [overlap analysis]]\n\n## 5. Ecosystem-Wide Contract Integrity Report\n[Calls cycles: [count] total, [list pairs], [hub nodes involved]\nImplements cycles: [count] total, [list pairs], [hub nodes involved]\nExtends cycles: [count] total, [list pairs], [hub nodes involved]\nUses cycles: [count] total, [list pairs], [hub nodes involved]\nImports cycles: [count] total, [list pairs], [hub nodes involved]\nCycle length distribution: [histogram]\nInterconnected cycle clusters: [analysis]]\n\n## 6. Deep Execution Flow Maps\n[For top 10 APIs by Ca:\nAPI: [id] (Ca=[value])\n- Call chain (depth 8): [complete path tree]\n- Execution depth: max=[value], mean=[value]\n- Shared bottleneck nodes: [list]\n- Leaf destinations: [count]\n- Path overlap with other APIs: [percentage]]\n\n## 7. Complete Interface/Trait Ecosystem\n[For each interface:\nInterface: [id]\n- Direct implementers: [count]\n- Transitive implementers (depth 4): [count by depth]\n- Inheritance chain (Extends, depth 4): [depth, ancestors]\n- Dependency fan-out: [metric]]\n\n## 8. API Boundary Crossing Analysis\n[For each API:\n- Module containment (depth 3): [module hierarchy]\n- Crosses external boundary: [yes/no]\n- Boundary crossing count: [metric]\nExternal-facing APIs: [count and list]]\n\n[Report ONLY factual graph measurements - NO interpretations]",
-  "endpoints": [
-    {
-      "name": "APIEndpoint",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42,
-      "api_type": "HTTP",
-      "description": "Brief description",
-      "dependencies": ["dep1", "dep2"]
-    }
-  ],
-  "usage_patterns": ["Pattern 1: Usage description"],
-  "integration_points": ["Integration point 1"]
-}
-MANDATORY: endpoints array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "endpoints": [{"name": "X", "file_path": "a.rs", "line_number": 1, "api_type": "HTTP", "description": "...", "dependencies": []}], "usage_patterns": [], "integration_points": []}
 
 CRITICAL RULES:
 1. Extract ALL node IDs from previous tool results - never invent them

crates/codegraph-mcp/src/architecture_analysis_prompts.rs

@@ -20,32 +20,9 @@ ANALYSIS FOCUS (report WHAT IS, not what SHOULD BE):
 - Architectural Patterns: Identify hub patterns, dependency structures
 - Code Smells: Detect circular dependencies, god objects (nodes with extreme hub degree)
 
-RESPONSE FORMAT (JSON):
-{
-  "reasoning": "Your analysis plan or findings",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { tool parameters from previous results }
-  },
-  "is_final": false
-}
-
-When complete, respond with ONLY this JSON (no 'reasoning', 'tool_call', or 'is_final' fields):
-{
-  "analysis": "ARCHITECTURE SUMMARY:\n- Coupling metrics: [data]\n- Circular deps: [count]\n- Hub nodes: [list with degrees]\n- Interpretation: [objective observations only]",
-  "layers": [],
-  "hub_nodes": [
-    {
-      "name": "HubNode",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42
-    }
-  ],
-  "coupling_metrics": [],
-  "patterns": [],
-  "issues": []
-}
-MANDATORY: hub_nodes array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "layers": [], "hub_nodes": [{"name": "X", "file_path": "a.rs", "line_number": 1}], "coupling_metrics": [], "patterns": [], "issues": []}
 
 CONSTRAINTS:
 - MAX 5 STEPS - prioritize key architectural indicators
@@ -76,32 +53,9 @@ ANALYSIS DIMENSIONS (report WHAT IS, not what SHOULD BE):
 - Code Smells: Detect circular dependencies, god objects, coupling hotspots
 - Change Impact: Assess blast radius using reverse dependencies
 
-RESPONSE FORMAT (JSON):
-{
-  "reasoning": "Your analysis plan or findings with specific metrics",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { tool parameters from previous results }
-  },
-  "is_final": false
-}
-
-When complete, respond with ONLY this JSON (no 'reasoning', 'tool_call', or 'is_final' fields):
-{
-  "analysis": "ARCHITECTURE ANALYSIS SUMMARY:\n\n## Coupling Metrics\n- Ca, Ce, I distributions for key nodes\n- Stability zones (stable I<0.3, unstable I>0.7)\n\n## Architectural Hotspots\n- Hub nodes with degrees and coupling metrics\n- God object candidates (high degree + high Ce)\n\n## Dependency Health\n- Circular dependencies count by edge type\n- Problematic dependency patterns\n\n## Change Impact Assessment\n- High-impact nodes (high Ca)\n- Blast radius for critical components\n\n## Interpretation\n[Objective observations only - no prescriptive judgments]",
-  "layers": ["Layer 1: Infrastructure", "Layer 2: Core"],
-  "hub_nodes": [
-    {
-      "name": "HubNode",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42
-    }
-  ],
-  "coupling_metrics": ["Node X: Ca=15, Ce=8, I=0.35"],
-  "patterns": ["Pattern: Layered architecture"],
-  "issues": ["Issue: Circular dependency between X and Y"]
-}
-MANDATORY: hub_nodes array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "layers": [], "hub_nodes": [{"name": "X", "file_path": "a.rs", "line_number": 1}], "coupling_metrics": [], "patterns": [], "issues": []}
 
 CONSTRAINTS:
 - MAX 10 STEPS - cover key architectural dimensions systematically
@@ -132,32 +86,9 @@ AVAILABLE TOOLS:
 5. get_reverse_dependencies(node_id, edge_type, depth) - Maps dependents (depth 3-5)
 6. trace_call_chain(from_node, max_depth) - Maps execution paths (depth 4-6)
 
-RESPONSE FORMAT (JSON):
-{
-  "reasoning": "Detailed analysis of current step with specific metrics and connections to previous findings",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { precise parameters based on prior results }
-  },
-  "is_final": false
-}
-
-When complete, respond with ONLY this JSON (no 'reasoning', 'tool_call', or 'is_final' fields):
-{
-  "analysis": "[Detailed architectural analysis with coupling distributions, hub analysis, dependency health, change impact, execution patterns, and objective synthesis]",
-  "layers": ["Layer 1: Infrastructure", "Layer 2: Core", "Layer 3: Application"],
-  "hub_nodes": [
-    {
-      "name": "HubNode",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42
-    }
-  ],
-  "coupling_metrics": ["Node X: Ca=15, Ce=8, I=0.35"],
-  "patterns": ["Pattern: Layered architecture"],
-  "issues": ["Issue: Circular dependency between X and Y"]
-}
-MANDATORY: hub_nodes array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "layers": [], "hub_nodes": [{"name": "X", "file_path": "a.rs", "line_number": 1}], "coupling_metrics": [], "patterns": [], "issues": []}
 
 CONSTRAINTS:
 - MAX 15 STEPS - thorough multi-dimensional analysis
@@ -184,32 +115,9 @@ AVAILABLE TOOLS (use extensively):
 5. get_reverse_dependencies(node_id, edge_type, depth) - Complete impact (depth 5-8)
 6. trace_call_chain(from_node, max_depth) - Exhaustive execution (depth 7-10)
 
-RESPONSE FORMAT (JSON):
-{
-  "reasoning": "Comprehensive step analysis with phase, objective, rationale, connections, pattern investigation, and coverage assessment",
-  "tool_call": {
-    "tool_name": "name_of_tool",
-    "parameters": { precise parameters extracted from results }
-  },
-  "is_final": false
-}
-
-When complete, respond with ONLY this JSON (no 'reasoning', 'tool_call', or 'is_final' fields):
-{
-  "analysis": "[Exhaustive architectural analysis with complete coupling landscape, architectural topology, dependency health, change impact, execution analysis, cross-dimensional patterns, and statistical health metrics]",
-  "layers": ["Layer 1: Infrastructure", "Layer 2: Core", "Layer 3: Application", "Layer 4: Presentation"],
-  "hub_nodes": [
-    {
-      "name": "HubNode",
-      "file_path": "relative/path/to/file.rs",
-      "line_number": 42
-    }
-  ],
-  "coupling_metrics": ["Node X: Ca=15, Ce=8, I=0.35"],
-  "patterns": ["Pattern: Layered architecture", "Pattern: Hub-and-spoke"],
-  "issues": ["Issue: Circular dependency between X and Y", "Issue: God object Z"]
-}
-MANDATORY: hub_nodes array must include file paths from tool results
+FORMAT:
+- Intermediate: {"reasoning": "...", "tool_call": {...}, "is_final": false}
+- Final: {"analysis": "...", "layers": [], "hub_nodes": [{"name": "X", "file_path": "a.rs", "line_number": 1}], "coupling_metrics": [], "patterns": [], "issues": []}
 
 CRITICAL RULES:
 1. ZERO HEURISTICS: Every claim must be based on tool output data