Skill

prompt-section-design

Design composable prompt sections for building agentic prompts. Use when creating reusable prompt components, designing LEGO-block prompt sections, or structuring prompts for the stakeholder trifecta.

From tac
Install
1
Run in your terminal
$
npx claudepluginhub melodic-software/claude-code-plugins --plugin tac
Tool Access

This skill is limited to using the following tools:

ReadGrepGlob
Skill Content

Prompt Section Design Skill

Design composable prompt sections that work like LEGO blocks for building prompts at any level.

Purpose

Create well-structured prompt sections that are reusable, consistent, and effective for the stakeholder trifecta.

When to Use

  • Creating a new prompt
  • Restructuring existing prompt
  • Adding sections to prompt
  • Standardizing team prompts

Section Tier List

TierSectionsPriority
SWorkflowAlways include (Level 2+)
AVariables, Examples, Control Flow, Delegation, TemplateHigh value
BPurpose, High-Level, Higher Order, InstructionsSupporting
CMetadata, Codebase Structure, Relevant Files, ReportAs needed

Design Process

Step 1: Identify Prompt Purpose

Ask:

  • What does this prompt accomplish?
  • Who will use it? (you, team, agents)
  • What level is it? (1-7)
  • What inputs/outputs are needed?

Step 2: Select Required Sections

Based on level:

LevelRequiredRecommended
1Title, Prompt-
2Title, WorkflowVariables, Report
3Title, WorkflowVariables, Control Flow
4Title, WorkflowVariables, Delegation
5Title, WorkflowVariables
6Title, Workflow, TemplateVariables
7Title, Workflow, ExpertiseVariables

Step 3: Design Each Section

Metadata (Frontmatter)

---
description: Clear, searchable description
argument-hint: [arg1] [arg2]
allowed-tools: Read, Write, Edit
model: sonnet
---

Guidelines:

  • description: What does it do? When to use?
  • argument-hint: What parameters expected?
  • allowed-tools: Minimal set needed
  • model: Match to task complexity

Title

# Action-Oriented Title

Guidelines:

  • Use imperative verb: Create, Build, Generate, Analyze
  • Be specific: "Create Implementation Plan" not "Plan"
  • Keep concise: 2-5 words

Purpose

## Purpose
[1-2 sentences describing what the prompt accomplishes]

Guidelines:

  • Direct language to agent
  • Reference key sections
  • Explain the "what" and "why"

Variables

## Variables

# Dynamic (from user)
USER_PROMPT: $ARGUMENTS
FILE_PATH: $1
COUNT: $2 or 3 if not provided

# Static (fixed)
OUTPUT_DIR: specs/
MODEL: sonnet

Guidelines:

  • SCREAMING_SNAKE_CASE
  • Dynamic first, static second
  • Include defaults where appropriate
  • Clear descriptions

Workflow (S-Tier)

## Workflow

1. Validate inputs
   - Check USER_PROMPT is provided
   - If not, STOP and ask user
2. Process task
   - Sub-step detail
3. Generate output
4. Report results

Guidelines:

  • Numbered steps for sequence
  • Sub-bullets for details
  • STOP conditions explicit
  • Clear progression

Instructions

## Instructions

- IMPORTANT: Always validate before processing
- Handle edge cases gracefully
- Never modify files outside project

Guidelines:

  • Bullet points for rules
  • IMPORTANT markers for critical
  • Edge cases explicit

Report

## Report

## Task Complete

**Files:** [count]
**Status:** [status]

### Changes
- [change 1]
- [change 2]

Guidelines:

  • Template for output
  • Consistent format
  • Easy to parse

Template (Level 6)

<!-- markdownlint-disable MD033 MD025 MD003 MD040 -->
## Specified Format

```text
---

allowed-tools: <tools>
description: <description>
---

# <name>

## Variables

<VAR>: $1

## Workflow

<steps>
<!-- markdownlint-enable MD033 MD025 MD003 MD040 -->

Guidelines:

  • Complete template
  • Placeholders marked clearly
  • Follows prompt conventions

Expertise (Level 7)

## Expertise

### Domain Knowledge
- Pattern 1 learned
- Pattern 2 discovered

### Discovered Patterns
- Implementation insight 1
- Best practice 2

Guidelines:

  • Organized by category
  • Grows over time
  • Never modify Workflow

Step 4: Validate Structure

Checklist:

  • Title is action-oriented
  • Workflow has numbered steps
  • Variables use SCREAMING_SNAKE_CASE
  • STOP conditions are explicit
  • Frontmatter has description
  • Sections in logical order

Section Order Convention

<!-- markdownlint-disable MD033 MD025 MD040 -->
---

[frontmatter]
---

# [Title]

## Purpose

[purpose]

## Variables

[variables]

## Instructions

[instructions]

## Workflow

[workflow]

## Report

[report format]
<!-- markdownlint-enable MD033 MD025 MD040 -->

Output Format

When designing sections:

## Section Design

**Prompt:** [name]
**Level:** [1-7]

### Recommended Sections

1. **Title**: [suggested title]

2. **Frontmatter**:

```yaml
description: ...
argument-hint: ...
allowed-tools: ...
model: ...
```

1. **Variables**:
   - Dynamic: [list]
   - Static: [list]

2. **Workflow**: [step count] steps
   - Step 1: [overview]
   - Step 2: [overview]
   ...

3. **Report**: [format type]

Anti-Patterns

Anti-PatternProblemSolution
No Workflow sectionAgent lacks directionAlways add for Level 2+
Inconsistent variable namesConfusionSCREAMING_SNAKE_CASE
Missing STOP conditionsRunaway executionExplicit early exits
Over-detailed workflowReduces agent autonomyHigh-level steps
No frontmatterHard to discoverAdd description

Key Quote

"Build libraries of reusable battle-tested agentic prompts with composable sections that work like LEGO blocks."

Cross-References

  • @prompt-sections-reference.md - Section definitions
  • @seven-levels.md - Sections by level
  • @variable-patterns.md - Variable conventions

Version History

  • v1.0.0 (2025-12-26): Initial release

Last Updated

Date: 2025-12-26 Model: claude-opus-4-5-20251101

Stats
Parent Repo Stars40
Parent Repo Forks6
Last CommitDec 27, 2025