Qlik Predict Skill for Claude

A Claude Code skill that guides users through the full machine learning lifecycle using Qlik Predict (AutoML) on Qlik Cloud — from use-case ideation through to model deployment and monitoring. It can also generate professional, data-driven ML Enhancement Strategy reports as shareable HTML documents.

---

What It Does

The skill covers four phases of a Qlik Predict project:

Phase	Description
Ideation & Use-Case Discovery	Suggests concrete ML questions based on columns, sample data, or industry context. Facilitates structured workshop brainstorming.
Problem Definition & Refinement	Narrows to a well-defined ML question with a clear grain, target, and experiment type.
Data Preparation & Feature Engineering	Provides working Qlik load script and Data Flow examples for aggregation, lag/window features, RFM, date engineering, class imbalance handling, and leakage prevention.
Model Evaluation & Iterative Improvement	Interprets scores, permutation importance, and SHAP values. Guides versioning, intelligent model optimization, deployment, and drift monitoring.

It also generates ML Enhancement Strategy reports — comprehensive HTML documents with executive summaries, prioritized use cases, feature engineering details with real Qlik code, business value calculations, and an implementation roadmap.

---

Files

├── README.md
├── .gitignore
└── qlik-predict/                         # Skill folder — zip this directory to install
    ├── SKILL.md                          # Skill definition and workflow routing
    └── references/
        ├── use-case-patterns.md          # Industry use-case library, workshop guide, forecasting dual-approach
        ├── data-prep-feature-engineering.md  # Qlik load script and Data Flow examples
        ├── model-evaluation-improvement.md   # Metrics, iterative refinement, deployment guidance
        ├── ml-strategy-report-generation.md  # HTML report structure, styling, quality checklist
        └── qlik-brand-colors-reference.md    # Official Qlik color palette and usage rules

---

Requirements

• Claude, Claude Code, or other AI platform with skills support
• Optional: Qlik MCP Server — enables the skill to query live Qlik apps and generate fully data-driven reports using real field names and actual data patterns

The skill works without MCP tools, but report generation will be based on the fields and data you describe rather than live app queries.

---

Installation

Claude Pro / Team / Enterprise (claude.ai or Claude Desktop)

Prerequisite: In Settings > Capabilities, ensure Code execution and file creation is enabled. On Enterprise plans, an organisation owner must enable this in Organisation settings > Capabilities first.

Upload the skill:

1. Download or clone this repository. Zip the qlik-predict folder — the ZIP must contain the folder (not just its contents), and the folder name must match the skill name (qlik-predict). README.md and .gitignore live in the repo root and are not part of the skill.
2. In Claude, go to Settings > Capabilities.
3. Scroll to the Skills section and click Upload skill.
4. Upload the ZIP file. The skill will appear in your Skills list and can be toggled on or off.

Uploaded skills are private to your account. On Team or Enterprise plans, organisation owners can provision skills to all members via Organisation settings.

Alternative — claude.ai Projects (manual setup, no ZIP needed):

If you prefer not to upload a ZIP, you can configure the skill manually inside a Project:

1. Create a new Project in Claude (e.g., "Qlik Predict Advisor").
2. Open Project instructions and paste in the contents of qlik-predict/SKILL.md (omit the --- frontmatter block at the top — that is only used by Claude Code).
3. In Project knowledge, upload the five files from qlik-predict/references/:
   • use-case-patterns.md
   • data-prep-feature-engineering.md
   • model-evaluation-improvement.md
   • ml-strategy-report-generation.md
   • qlik-brand-colors-reference.md
4. Start a conversation in the project and use any of the prompts in the Usage section below.

Claude Code

Copy the qlik-predict folder (found in the repo root) into your Claude skills directory:

# macOS / Linux
cp -r qlik-predict ~/.claude/skills/

# Windows
xcopy /E /I "qlik-predict" "%USERPROFILE%\.claude\skills\qlik-predict"

Claude Code will automatically load the skill and trigger it based on context.

---

Usage

The skill is triggered automatically when Claude detects relevant context, or you can invoke it explicitly in your prompt.

Ideation — "What ML could we do with this data?"

I have a dataset with columns: CustomerID, OrderDate, ProductCategory, Amount, Region, Churned.
What ML use cases could Qlik Predict support?

We're a manufacturing company tracking equipment sensor readings and maintenance history.
Use the Qlik Predict skill to suggest ML use cases for our data.

Workshop facilitation

I'm running a workshop with our analytics team to brainstorm ML opportunities.
Can you facilitate a structured session using the Qlik Predict skill?
We're in the logistics industry.

Data prep and feature engineering

I'm building a churn prediction model in Qlik Predict.
My source data is transactional (one row per order). Help me prepare the dataset.

Model evaluation and improvement

My Qlik Predict binary classification model has F1 = 0.61 and one feature dominates
the permutation importance chart at 85%. How should I improve it?

ML Strategy Report — with Qlik MCP Server

When connected to a live Qlik app via the Qlik MCP Server, the skill queries actual field names, data volumes, and patterns before writing any content:

Can you use the Qlik Predict skill to consider how ML can be applied to this dataset?
Present the results as a report.

Analyze the Safety Compliance Inspections app and then use the Qlik Predict skill
to recommend ML use cases that could be based on this dataset.
Use the guidance in the skill to generate a comprehensive report.

ML Strategy Report — without MCP tools

Paste column names, a data dictionary, or sample rows and ask for a report directly:

Here are the fields in my dataset: [paste field list].
Use the Qlik Predict skill to generate an ML Enhancement Strategy report.

---

Report Output

ML Enhancement Strategy reports are generated as self-contained HTML files. They include:

• Executive summary with headline metrics
• Data model overview with field-level detail
• 3–5 prioritized ML use cases, each with: ML question, feature engineering, working Qlik code, business value calculation, and recommended actions
• Implementation roadmap (phased, with realistic timeframes)
• Data preparation requirements with load script examples
• Total business value summary

Reports are formatted for executive presentation or printing.

---

Disclaimer

This skill uses AI to generate ML recommendations, Qlik load script code, feature engineering guidance, and business value estimates. All output should be treated as a starting point and reviewed by a qualified person before use.

In particular:

• Qlik code examples should be tested in a development environment before use in production. Field names, table names, and expressions must be verified against your actual data model.
• Business value projections are illustrative estimates based on general assumptions. Validate against your own data, costs, and operational context.
• ML feasibility assessments depend on data quality, volume, and characteristics that the AI cannot fully evaluate without access to the actual data.
• Feature engineering and leakage guidance is based on general best practices. Domain expertise is required to confirm that features are genuinely available at prediction time in your specific context.

AI can and does make mistakes. Do not deploy models, load scripts, or business cases based solely on this output without independent review.