# Development Modes for Chapters in the Presentation Skill: A, B, and C Modes Explained

> Explore presentation skill development modes A B and C in ConardLi/garden-skills. Learn about per-chapter review sequential batch processing and parallel sub-agent execution for effective chapter development.

- Repository: [ConardLi/garden-skills](https://github.com/ConardLi/garden-skills)
- Tags: deep-dive
- Published: 2026-08-29

---

**The Web‑Video‑Presentation skill in ConardLi/garden‑skills offers three distinct development modes for chapters after the initial anchor: Mode A (逐章确认) for per‑chapter review, Mode B (第1章后顺序开发) for sequential batch processing, and Mode C (第1章后并行开发) for parallel sub‑agent execution.**

The **Web‑Video‑Presentation** skill in the ConardLi/garden‑skills repository enables developers to build click‑driven 16:9 web demos that simulate video presentations. Understanding the **development modes for chapters in the presentation skill** is critical for optimizing your workflow, as each mode offers distinct trade‑offs between quality assurance, execution speed, and resource utilization. After the first chapter—which always serves as the anchor—you can select how the remaining chapters are generated according to the **Checkpoint Plan** defined in [`skills/web-video-presentation/SKILL.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/SKILL.md).

## The Three Development Modes for Chapter Generation

According to the source code in [`skills/web-video-presentation/SKILL.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/SKILL.md), the workflow explicitly lists three options that determine how agents schedule chapter generation. All modes rely on the same hard rules defined in [`skills/web-video-presentation/references/CHAPTER-CRAFT.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/references/CHAPTER-CRAFT.md), but differ in execution strategy and review cadence.

### Mode A: 逐章确认 (Per-Chapter Review)

**Mode A** is the default and safest approach. In this workflow, the agent builds each chapter individually and then pauses for user‑led acceptance before proceeding to the next. This **per‑chapter review** process catches visual or timing issues early, preventing compounding errors across the presentation.

Use this mode when you require the highest quality guarantee or when the agent cannot parallelize work effectively. The checkpoint after every chapter ensures that [`narrations.ts`](https://github.com/ConardLi/garden-skills/blob/main/narrations.ts) constraints and visual principles from [`CHAPTER-CRAFT.md`](https://github.com/ConardLi/garden-skills/blob/main/CHAPTER-CRAFT.md) are strictly verified before continuation.

### Mode B: 第1章后顺序开发 (Sequential After Chapter 1)

**Mode B** initiates sequential development immediately after the first anchor chapter is accepted. Rather than pausing between chapters, the agent creates the remaining chapters one after another without intermediate stops. All chapters complete before a single collective review occurs.

This mode suits environments where speed matters but parallel execution is not supported. While faster than Mode A due to reduced context switching, it delays quality verification until the entire sequence finishes.

### Mode C: 第1章后并行开发 (Parallel with Sub-Agents)

**Mode C** represents the fastest execution path. After the anchor chapter, subsequent chapters dispatch to separate sub‑agents that build content in parallel. You control the parallelism level—for example, running 4 chapters simultaneously—through configuration parameters.

While this maximizes throughput, the source code notes that visual styles may differ between chapters. This variation is often intentional, mimicking multi‑voice video production. Choose this mode when you prioritize maximal speed and can manage sub‑agents, accepting possible stylistic variation across the presentation.

## Implementing Development Modes Programmatically

You can select and execute these **development modes for chapters in the presentation skill** through TypeScript logic or CLI invocation. The following examples demonstrate how to parse user selection and handle the generation loop accordingly.

### TypeScript Implementation

The enum definition and switch logic map user input to the specific workflow:

```typescript
// Enumeration of available development modes
enum DevMode {
  PerChapter = 'A',   // 逐章确认
  Sequential = 'B',   // 第1章后顺序开发
  Parallel   = 'C',   // 第1章后并行开发
}

// Parse user selection with fallback to safest mode
function getDevMode(userChoice: string): DevMode {
  switch (userChoice.toUpperCase()) {
    case 'A': return DevMode.PerChapter;
    case 'B': return DevMode.Sequential;
    case 'C': return DevMode.Parallel;
    default: return DevMode.PerChapter; // Default to safest mode
  }
}

// Execute generation based on selected mode
const mode = getDevMode(userChoice);
for (let i = 2; i <= totalChapters; i++) {
  if (mode === DevMode.PerChapter) {
    await buildChapter(i);
    await awaitUserAcceptance(i);
  } else if (mode === DevMode.Sequential) {
    await buildChapter(i);
  } else { // Parallel mode
    launchSubAgent(i); // Fire-and-forget; sub-agent handles acceptance
  }
}

```

### CLI Usage Examples

Initialize your presentation and specify the mode using the scaffold script:

```bash

# List available themes first

bash scripts/scaffold.sh --list-themes

# Scaffold a new presentation with selected theme

bash scripts/scaffold.sh ./presentation --theme=my-favorite-theme

# After chapter 1 completion, specify development mode:

#   A – per-chapter review

#   B – sequential development

#   C – parallel (specify max parallelism, e.g., 4)

```

## Key Configuration Files and Their Roles

Understanding the **development modes for chapters in the presentation skill** requires familiarity with these specific files in the ConardLi/garden‑skills repository:

- **[`skills/web-video-presentation/SKILL.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/SKILL.md)** — Contains the full workflow description, including the *Checkpoint Plan* section that defines the three modes (A, B, and C) and notes that users can switch modes after chapter 1 is accepted.

- **[`skills/web-video-presentation/references/CHAPTER-CRAFT.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/references/CHAPTER-CRAFT.md)** — Defines the hard rules each chapter must obey, including step counting, [`narrations.ts`](https://github.com/ConardLi/garden-skills/blob/main/narrations.ts) constraints, and visual principles. All development modes rely on this single source of truth for chapter implementation.

- **[`scripts/scaffold.sh`](https://github.com/ConardLi/garden-skills/blob/main/scripts/scaffold.sh)** — Provides the command‑line entry point for initializing new presentation projects and selecting themes, which precedes the mode selection phase.

- **[`skills/web-video-presentation/references/THEMES.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/references/THEMES.md)** — Lists built‑in themes and token contracts; the chosen theme influences how each development mode renders visual consistency, particularly impacting Mode C where style variations may occur.

## Summary

- **Mode A (逐章确认)** offers the safest workflow with per‑chapter user acceptance, ideal for high‑quality requirements.
- **Mode B (第1章后顺序开发)** accelerates delivery by processing chapters sequentially without intermediate pauses, suitable for speed‑focused sequential environments.
- **Mode C (第1章后并行开发)** maximizes speed through sub‑agent parallelization but may introduce intentional stylistic variations between chapters.
- All modes require completion of the first anchor chapter before activation and can be switched dynamically after chapter 1 acceptance.
- The underlying chapter implementation remains consistent across modes, governed by [`CHAPTER-CRAFT.md`](https://github.com/ConardLi/garden-skills/blob/main/CHAPTER-CRAFT.md) rules regardless of execution strategy.

## Frequently Asked Questions

### Can I switch development modes after completing the first chapter?

Yes. According to the **Checkpoint Plan** in [`skills/web-video-presentation/SKILL.md`](https://github.com/ConardLi/garden-skills/blob/main/skills/web-video-presentation/SKILL.md), you can switch modes at any point after chapter 1 is accepted. This flexibility allows you to start with the safer Mode A for initial validation, then switch to Mode B or C for faster bulk generation of remaining chapters.

### Will parallel mode (Mode C) create inconsistent visual styles between chapters?

Yes, this is an intentional possibility. When using **Mode C (第1章后并行开发)**, different sub‑agents may produce varying visual styles across chapters. The source code explicitly notes this mimics multi‑voice video production. If strict visual consistency is required, use Mode A or B instead.

### What is the default development mode if I do not specify one?

The system defaults to **Mode A (逐章确认)**. The `getDevMode` function in the reference implementation returns `DevMode.PerChapter` for any unrecognized or undefined input, ensuring the safest workflow is active by default.

### How does the first chapter influence the development modes?

The first chapter always serves as the **anchor** chapter across all modes. You must complete and accept this initial chapter before any development mode (A, B, or C) activates for subsequent chapters. This anchor establishes the baseline [`narrations.ts`](https://github.com/ConardLi/garden-skills/blob/main/narrations.ts) structure and visual theme that guides all following content, whether processed sequentially or in parallel.