Marketplace

prompt-level-selection

Guide selection of appropriate prompt level for a task. Use when choosing between simple prompts and complex workflows, applying the seven levels framework, or matching task complexity to prompt investment.

allowed_tools: Read, Grep, Glob

$ インストール

git clone https://github.com/melodic-software/claude-code-plugins /tmp/claude-code-plugins && cp -r /tmp/claude-code-plugins/plugins/tac/skills/prompt-level-selection ~/.claude/skills/claude-code-plugins

// tip: Run this command in your terminal to install the skill


name: prompt-level-selection description: Guide selection of appropriate prompt level for a task. Use when choosing between simple prompts and complex workflows, applying the seven levels framework, or matching task complexity to prompt investment. allowed-tools: Read, Grep, Glob

Prompt Level Selection Skill

Guide selection of the appropriate prompt level for a task using the seven levels framework.

Purpose

Match task complexity to the right prompt level. Start simple, add complexity only when needed.

When to Use

  • Starting a new prompt
  • Upgrading existing prompt
  • Unsure which level fits
  • Teaching prompt engineering

The Seven Levels Quick Reference

LevelNameUse When
1High-LevelSimple repeatable task
2WorkflowSequential steps needed
3Control FlowConditionals or loops
4DelegationMultiple agents needed
5Higher-OrderProcessing other prompts
6Template MetaGenerating prompts
7Self-ImprovingKnowledge accumulation

Decision Tree

Is this a simple, repeatable task?
├── Yes -> Level 1 (High-Level Prompt)
└── No
    |
    Does it need sequential steps?
    ├── Yes -> Does it need conditionals/loops?
    |   ├── Yes -> Does it delegate to agents?
    | |   ├── Yes -> Level 4 (Delegation)
    | |   └── No -> Level 3 (Control Flow)
    |   └── No -> Level 2 (Workflow)
    └── No
        |
        Does it process other prompts?
        ├── Yes -> Level 5 (Higher-Order)
        └── No
            |
            Does it generate prompts?
            ├── Yes -> Level 6 (Template Meta)
            └── No
                |
                Does it need to learn over time?
                ├── Yes -> Level 7 (Self-Improving)
                └── No -> Reassess requirements

The 80/20 Rule

"Levels 3-4 cover 80% of practical use cases."

Level RangeCoverageComplexity
1-240%Low
3-480%Medium (Sweet Spot)
5-720%High

Don't over-engineer. Most tasks fit in Levels 3-4.

Selection Process

Step 1: Understand the Task

Ask:

  • What does this task accomplish?
  • How often will it be repeated?
  • What inputs does it need?
  • What outputs does it produce?

Step 2: Check Complexity Indicators

IndicatorPoints To
One-time or rareLevel 1
Sequential stepsLevel 2+
"If X then Y"Level 3+
"Run N agents"Level 4
"Process this spec file"Level 5
"Create prompts for..."Level 6
"Learn and improve"Level 7

Step 3: Start Low, Upgrade If Needed

  1. Start with lowest applicable level
  2. Build and test the prompt
  3. If insufficient, upgrade one level
  4. Repeat until task is satisfied

Level Selection Examples

Example 1: "Start the dev server"

Analysis: Simple, repeatable, no variables Level: 1 (High-Level Prompt)

Example 2: "Create an implementation plan"

Analysis: Sequential steps, needs input, produces output Level: 2 (Workflow Prompt)

Example 3: "Generate N images with validation"

Analysis: Loop required, conditional checking Level: 3 (Control Flow)

Example 4: "Research topic with 5 parallel agents"

Analysis: Multiple agents, aggregation needed Level: 4 (Delegation Prompt)

Example 5: "Build from this spec file"

Analysis: Accepts another prompt/spec as input Level: 5 (Higher-Order)

Example 6: "Create new slash commands"

Analysis: Generates prompts in specific format Level: 6 (Template Meta Prompt)

Example 7: "Hook expert that learns patterns"

Analysis: Accumulates expertise over time Level: 7 (Self-Improving Prompt)

Output Format

When recommending a level:

## Level Selection

**Task:** [description]

**Recommended Level:** [1-7] ([name])

**Rationale:**
- [reason 1]
- [reason 2]

**Key Sections Needed:**
- [section 1]
- [section 2]

**Alternative Consideration:**
Level [N] if [condition]

Red Flags

Red FlagIssueSolution
Jumping to Level 6-7Over-engineeringStart at Level 2-3
Level 1 for complex taskUnder-engineeringAdd Workflow section
Level 4 for single-agentUnnecessary delegationUse Level 2-3
No clear level fitVague requirementsClarify task scope

Key Quote

"Three times marks a pattern. Copy whatever you're doing and write it as a high level prompt, then move up the levels from there."

Cross-References

  • @seven-levels.md - Detailed level descriptions
  • @prompt-sections-reference.md - Sections for each level
  • @stakeholder-trifecta.md - Communication considerations

Version History

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

Last Updated

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