Prompt Complexity Scorer

Prompt Complexity Scorer

This skill evaluates user prompts to determine if they require planning before implementation.

Skip Evaluation

Do not evaluate prompts that are explicit slash command invocations. If the user's prompt starts with /, they are already invoking a workflow and no complexity scoring is needed.

Examples of prompts to skip (just execute the command):

• /​project:bootstrap @specs/​add-auth.md
• /​git:commit
• /​project:implement @projects/​2026-01-26-feature

Examples of prompts to evaluate:

• "Add WebSocket support to this project"
• "Make the app faster"
• "Refactor the authentication system"

Scoring Criteria

Score each prompt on a 1-10 scale based on these factors:

Score Calculation

1. Evaluate each factor independently
2. Average the factor scores
3. Round to nearest integer

Behavior by Score

Score 1-4: Continue Normally

Do not mention the scoring. Proceed with the request immediately.

Score 5-10: Suggest Project

Pause and ask the user:

This request scores [X]/​10 on complexity. I suggest creating a project to track this work properly.

Would you like me to create `projects/<date>-<suggested-name>/` with a brief.md?

Where <date> is today's date in YYYY-MM-DD format and <suggested-name> is a kebab-case name derived from the request (e.g., "2026-01-24-add-websockets", "2026-01-24-refactor-auth-system").

Project Setup

When creating a project, create the directory structure and brief.md:

# Get today's date
DATE=$(date +%Y-%m-%d)
mkdir -p projects/​${DATE}-<suggested-name>/​tasks

# <Title derived from prompt>

## Original Request

<User's exact prompt/​request>

## Goals

<Bullet points summarizing what needs to be accomplished>

## Notes

<Any additional context or constraints mentioned>

After creating the project, inform the user:

Project created at `projects/<date>-<suggested-name>/`.

You can now:
- Run `/​project:bootstrap @projects/<date>-<suggested-name>` for full research and planning
- Or continue with the request and tasks will be tracked in this project

IMPORTANT: After creating the project, set the active project context by creating a marker file:

echo "${DATE}-<suggested-name>" > .claude-active-project

This enables automatic task syncing to the project directory.

Examples

Example 1: Simple Request (Score 1)

Prompt: "Fix the typo in the error message on line 45 of user.service.ts"

Factors:

• Scope: 1 (single line)
• Clarity: 1 (exact location specified)
• Dependencies: 1 (none)
• Unknowns: 1 (none)
• Risk: 1 (trivial change)

Score: 1 → Continue normally

Example 2: Complex Request (Score 7)

Prompt: "Add WebSocket support to this project"

Factors:

• Scope: 8 (new system, multiple files)
• Clarity: 6 (what kind of WebSocket? real-time features?)
• Dependencies: 7 (infrastructure, client changes)
• Unknowns: 7 (architecture decisions needed)
• Risk: 7 (architectural change)

Score: 7 → Suggest creating projects/​YYYY-MM-DD-add-websockets/

Example 3: Medium Request (Score 3)

Prompt: "Add a new field 'nickname' to the User entity"

Factors:

• Scope: 4 (entity, migration, possibly resolvers)
• Clarity: 2 (clear requirement)
• Dependencies: 3 (migration, schema)
• Unknowns: 2 (standard pattern)
• Risk: 3 (database change but reversible)

Score: 3 → Continue normally

Example 4: Nebulous Request (Score 8)

Prompt: "Make the app faster"

Factors:

• Scope: 9 (entire application)
• Clarity: 9 (completely vague)
• Dependencies: 8 (unknown until investigated)
• Unknowns: 9 (what's slow? why?)
• Risk: 6 (depends on changes)

Score: 8 → Suggest creating projects/​YYYY-MM-DD-performance-optimization/

Important Notes

• This evaluation should be quick and silent for low-complexity requests
• Never mention the scoring system for scores 1-4
• For borderline cases (score 4-5), lean toward continuing normally
• The goal is to catch truly complex/​nebulous requests that benefit from planning