Appendix C Source Map: Which Files Ground Each Chapter

This book is not a full source walkthrough, but it is source-grounded. This appendix maps each chapter to its primary supporting files.

This is not a complete index. It lists only files directly relevant to the book's arguments.

One copyright boundary note: this map is for citing analytical basis, not for promising source-text reproduction. It preserves necessary engineering references, module positioning, and structural analysis, without shipping source copies or long implementation excerpts.

C.1 Chapter 1 Why Harness Engineering Matters

Core files:

• src/constants/prompts.ts
• src/utils/systemPrompt.ts
• src/query.ts
• src/services/tools/toolOrchestration.ts
• src/tools/BashTool/prompt.ts

Primary argument basis:

• prompt as control-plane component rather than personality packaging
• query loop as agent-system skeleton
• tool and Bash risk profiles as proof harness is necessary

C.2 Chapter 2 Prompt Is the Control Plane, Not Persona Decoration

Core files:

• src/constants/prompts.ts
• src/utils/systemPrompt.ts
• src/utils/claudemd.ts
• src/memdir/memdir.ts
• src/constants/systemPromptSections.ts
• src/main.tsx

Primary argument basis:

• layered system-prompt assembly
• CLAUDE.md and memory as control-plane inputs
• dynamic reminders and context injection

C.3 Chapter 3 Query Loop: The Heartbeat of an Agent System

Core files:

• src/query.ts
• src/QueryEngine.ts

Primary argument basis:

• state-machine properties of query loop
• input governance before model invocation
• streaming event consumption and recovery branches
• QueryEngine ownership of conversation lifecycle

C.4 Chapter 4 Tools, Permissions, and Interrupts

Core files:

• src/services/tools/toolOrchestration.ts
• src/services/tools/toolExecution.ts
• src/services/tools/StreamingToolExecutor.ts
• src/hooks/useCanUseTool.tsx
• src/utils/permissions/PermissionResult.ts
• src/tools/BashTool/prompt.ts
• src/tools/BashTool/bashPermissions.ts

Primary argument basis:

• concurrency safety and ordered context replay
• wrapped tool execution stack
• allow / deny / ask authorization semantics
• streaming-tool interruption and synthetic closure
• special high-pressure governance for Bash

C.5 Chapter 5 Context Governance: Memory, CLAUDE.md, and Compact

Core files:

• src/utils/claudemd.ts
• src/memdir/memdir.ts
• src/services/SessionMemory/prompts.ts
• src/services/compact/autoCompact.ts
• src/services/compact/compact.ts
• src/query.ts

Primary argument basis:

• layered discovery and loading of CLAUDE.md
• MEMORY.md as entry index, not content warehouse
• structured continuity in session memory
• autocompact thresholds, buffers, and circuit breaker
• post-compact reconstruction of working semantics

C.6 Chapter 6 Errors and Recovery

Core files:

• src/query.ts
• src/services/compact/autoCompact.ts
• src/services/compact/compact.ts
• src/services/api/withRetry.ts

Primary argument basis:

• withheld recoverable errors
• prompt-too-long handling via collapse drain and reactive compact
• max_output_tokens escalation and continuation strategy
• autocompact failure circuit breaking
• compaction self-recovery under PTL

C.7 Chapter 7 Multi-Agent and Verification

Core files:

• src/utils/forkedAgent.ts
• src/coordinator/coordinatorMode.ts
• src/tasks/LocalAgentTask/LocalAgentTask.tsx
• src/utils/hooks/hooksConfigManager.ts
• src/skills/bundled/verify.ts
• src/memdir/memoryTypes.ts

Primary argument basis:

• cache-safe fork parameters and state isolation
• coordinator synthesis responsibility
• independent verification stage
• subagent lifecycle hooks
• task cleanup and parent-child abort propagation
• stale-memory verification discipline

C.8 Chapter 8 Team Adoption

Core files:

• src/utils/claudemd.ts
• src/tools/SkillTool/prompt.ts
• src/tools/SkillTool/SkillTool.ts
• src/utils/forkedAgent.ts
• src/utils/hooks/hooksConfigManager.ts
• src/main.tsx

Primary argument basis:

• layered stability in team CLAUDE.md
• skills as institutional slices rather than prompt collections
• approval boundaries and allow-rule scoping
• lifecycle governance through hooks
• session startup behavior around instructions and skill loading

C.9 Chapter 9 Ten Principles

Chapter 9 is not derived from one single file. It compresses cross-chapter structures jointly reflected in:

• prompt assembly
• query loop
• tool orchestration
• permission model
• context governance
• recovery system
• multi-agent orchestration
• team governance hooks