Backmatter
Backmatter
Figure and Table Listing Policy
This file is the source of truth for building the figure list and table list in ebook and print backmatter. The figure sources for the English edition live under `manuscript-en/figures/`, and the table sources live in ea
This file is the source of truth for building the figure list and table list in ebook and print backmatter. The figure sources for the English edition live under manuscript-en/figures/, and the table sources live in each chapter's Markdown tables.
Role
- Help readers re-find figures and tables after they finish the book
- Preserve a stable seed before production adds figure numbers, table numbers, and page references
- Keep the figure plan, chapter prose, and production output aligned
Figure List Policy
- Treat
manuscript-en/figures/figure-plan.mdas the source of truth for figure ID, chapter, caption, and first mention - Keep each caption reader-facing so the main lesson of the figure is visible in one line
- Preserve the rule of one figure for one main message, then swap only the production numbering later
- Keep figure IDs aligned with the Japanese edition so cross-edition editorial work stays stable
Current Figure Seed
| Figure ID | Chapter | First Mention | Source |
|---|---|---|---|
fig-01 |
CH01 | The map from Prompt / Context / Harness | manuscript-en/figures/fig-01-maturity-model.mmd |
fig-02 |
CH05 | Persistent, task, session, and tool context | manuscript-en/figures/fig-02-context-classes.mmd |
fig-03 |
CH07 | The minimum input for resuming a session | manuscript-en/figures/fig-03-resume-packet.mmd |
fig-04 |
CH09 | The overall single-agent harness flow | manuscript-en/figures/fig-04-single-agent-harness.mmd |
fig-05 |
CH10 | The order of lint, typecheck, unit, and e2e | manuscript-en/figures/fig-05-verification-pipeline.mmd |
fig-06 |
CH11 | The split among planner, coder, and reviewer | manuscript-en/figures/fig-06-long-running-multi-agent.mmd |
fig-07 |
CH12 | The responsibilities that remain with humans | manuscript-en/figures/fig-07-operating-model.mmd |
Table List Policy
- Include only reader-facing tables, not temporary comparison tables or list substitutes
- Confirm production table titles by pairing the nearest heading with the table's main message
- Until page numbers exist, keep the seed in terms of return-to chapter and nearby heading
Current Table Seed
| Return-to Chapter | Nearby Heading | Table Role |
|---|---|---|
| CH01 | Getting one answer right is different from getting work done | Shows the difference between plausible output and completed work |
| CH02 | Separate objective, constraints, completion criteria, and forbidden actions | Shows the Prompt Contract elements and the failures caused by missing ones |
| CH04 | Evaluate prompts as operational artifacts | Shows the boundary among prompt, context, and harness evaluation |
| CH07 | Task context and session memory | Shows the minimum elements required for a restart |
| CH09 | Foundations of Harness Engineering | Compresses the flow from init to final report |
| CH10 | Verification is a system, not a single test run | Shows the order from failing test to approval |
| CH11 | Decide when to stay single-agent and when to split roles | Shows the branching conditions between single-agent and multi-agent work |
| CH12 | Operate by metrics, review budget, and hygiene | Shows the measurement group for throughput, quality, and hygiene |