dqe-quality: Data Quality Suite for Claude Code

5 skills for data quality directly inside Claude Code. Audit CSVs locally, or run a full end-to-end campaign data quality workflow — email validation, phone validation, deduplication — through the DQE One Server via MCP.

Note: the plugin is named dqe-quality in the marketplace. The GitHub repository is named claude-quality.

Drop a CSV and get a professional audit report. Or go further: connect to the DQE One Server and let Claude Code run the full data quality pipeline on your contact list — validation, deduplication, campaign-ready count — without leaving your terminal.

---

Table of Contents

• Who this is for
• Skills overview
• Installation
• Quick start
• dqe-audit — CSV audit report
• dqe-campaign — End-to-end campaign workflow
• dqe-list — Workspace overview
• dqe-dedup — Create a deduplication process
• dqe-run — Run a process
• DQE One Server — MCP setup
• The 6 dimensions
• The reports
• Options
• Output files
• File size handling
• Requirements
• License

---

Who this is for

Data engineers and analysts who need a fast, reproducible quality baseline on any CSV before loading it into a pipeline or CRM.

Marketing and campaign teams who need to clean a contact list fast — validate emails, fix phone numbers, remove duplicates — before launching a campaign.

Project managers and consultants who need ready-to-share deliverables — an audit report with actionable next steps and, optionally, an internal treatment plan — without opening a BI tool.

DQE Software teams who audit client data files and need branded, multilingual reports that tie findings directly to DQE service recommendations.

---

Skills overview

Skill	Invocation	Requires MCP	Description
dqe-audit	/dqe-quality:dqe-audit <file.csv>	No	Full 6-dimension audit → branded HTML report
dqe-campaign	/dqe-quality:dqe-campaign <file.csv>	Optional	Local audit + server-side email/phone/dedup → campaign-ready count
dqe-list	/dqe-quality:dqe-list	Yes	Workspace overview: processes, runs, files, rulesets
dqe-dedup	/dqe-quality:dqe-dedup	Yes	Guided deduplication process creation
dqe-run	/dqe-quality:dqe-run [name]	Yes	Trigger a process and follow its status

The three MCP skills (dqe-list, dqe-dedup, dqe-run) require a running DQE One Server instance configured as an MCP server. See DQE One Server — MCP setup.

---

Installation

Plugin install (recommended — CLI users)

/plugin install dqe-quality

The plugin is registered as dqe-quality in the marketplace. The underlying GitHub repository is DQE-SOFTWARE/claude-quality.

---

Desktop app install — Windows (no git required)

Open PowerShell and run:

irm https://raw.githubusercontent.com/DQE-SOFTWARE/claude-quality/main/install-desktop.ps1 | iex

This downloads the skill ZIP from GitHub, extracts it, and copies it to %USERPROFILE%\.claude\skills\. Restart Claude Code desktop when done.

Execution policy error? Run Set-ExecutionPolicy -Scope CurrentUser RemoteSigned first, then retry.

---

Desktop app install — macOS (no git required)

Open Terminal and run:

curl -fsSL https://raw.githubusercontent.com/DQE-SOFTWARE/claude-quality/main/install-desktop.sh | bash

This downloads the skill ZIP from GitHub, extracts it, and copies it to ~/.claude/skills/. Restart Claude Code desktop when done.

---

Manual install (CLI / Linux / git users)

git clone --depth 1 https://github.com/DQE-SOFTWARE/claude-quality.git
bash claude-quality/install.sh

---

Quick start

# Local audit — English HTML report
/dqe-quality:dqe-audit ~/data/contacts.csv

# Full campaign workflow (local audit + server cleaning + campaign-ready count)
/dqe-quality:dqe-campaign ~/data/landing-page-june.csv

# See your DQE One Server workspace
/dqe-quality:dqe-list

# Create a deduplication process (guided)
/dqe-quality:dqe-dedup

# Run an existing process
/dqe-quality:dqe-run "Father's Day campaign"

---

dqe-audit — CSV audit report

/dqe-quality:dqe-audit <path/to/file.csv> [--lang=fr|en|us|de|es] [--pm]

Runs a full data quality audit on a CSV file and generates 1 or 2 standalone HTML reports. Everything runs locally — your data never leaves your machine.

Argument	Description
path/to/file.csv	Path to the CSV file — relative, absolute, or Windows format
--lang=XX	Report language: en (default), us (alias for en), fr, de, es
--pm	Also generate the internal Project Manager guide (off by default)

# French report + PM guide
/dqe-quality:dqe-audit ~/data/clients.csv --lang=fr --pm

# Spanish report
/dqe-quality:dqe-audit ~/data/clientes.csv --lang=es

# Windows path (auto-converted to WSL)
/dqe-quality:dqe-audit "C:\Users\demo\data\export.csv" --lang=de

---

dqe-campaign — End-to-end campaign workflow

/dqe-quality:dqe-campaign [path/to/contacts.csv | --source=<type>] [--lang=fr|en] [--no-server]

The flagship skill. Audits your contact data and cleans it through the DQE One Server — email validation, phone validation, deduplication — producing a campaign-ready contact count.

The data can come from a local CSV file or directly from any remote source connected to the DQE One Server.

Data sources

Source	Invocation
Local CSV	/dqe-quality:dqe-campaign ~/data/contacts.csv
Salesforce	/dqe-quality:dqe-campaign --source=salesforce
Microsoft Dynamics 365	/dqe-quality:dqe-campaign --source=dynamics
PostgreSQL	/dqe-quality:dqe-campaign --source=postgres
Google BigQuery	/dqe-quality:dqe-campaign --source=bigquery
Snowflake	/dqe-quality:dqe-campaign --source=snowflake
SFTP	/dqe-quality:dqe-campaign --source=sftp
Interactive (ask me)	/dqe-quality:dqe-campaign

Flow — Local CSV

1. Local audit — detects email/phone columns, validates formats, counts duplicates, calculates a quality score
2. Inline audit card — score, key metrics, campaign-ready estimate
3. Server-side cleaning via DQE One Server (if MCP configured and confirmed):
   • Uploads the file, creates and runs check_email, check_phone, check_duplicate
4. Campaign-ready count

Flow — Remote source (Salesforce, Dynamics, PostgreSQL, BigQuery, Snowflake, SFTP)

1. Credential selection — lists available connections from list_credentials
2. Table / object selection — asks for the Salesforce object, Dynamics entity, SQL table, BigQuery dataset.table, Snowflake database.schema.table, or SFTP path
3. Field mapping — asks for primary key, email field, phone field
4. Server-side processes — creates and runs check_email, check_phone, check_duplicate directly on the remote data — no local file needed
5. Campaign-ready count

# Interactive — skill asks where the data is
/dqe-quality:dqe-campaign

# From a local CSV
/dqe-quality:dqe-campaign ~/data/landing-page-june.csv

# From Dynamics 365 (skip the source question)
/dqe-quality:dqe-campaign --source=dynamics

# From Salesforce
/dqe-quality:dqe-campaign --source=salesforce

# Local audit only — no MCP calls
/dqe-quality:dqe-campaign ~/data/contacts.csv --no-server

Example output (local CSV):

╔══════════════════════════════════════════════════════════════╗
║  DQE Data Quality Audit                                      ║
║  landing-page-fathers-day.csv · 1 247 records · 6 columns    ║
╠══════════════════════════════════════════════════════════════╣
║  Quality Score: 61/100  🟡                                   ║
╠════════════════════════════════════╦═════════════════════════╣
║  📧 Email validation               ║  📱 Phone validation    ║
║  Valid   : 1 089 (87%)             ║  Valid : 1 034 (83%)    ║
║  Invalid : 112                     ║  Invalid: 213           ║
║  Missing : 46                      ║  Missing: 0             ║
╠════════════════════════════════════╩═════════════════════════╣
║  👥 Duplicates detected: 89 records in 41 groups             ║
╠══════════════════════════════════════════════════════════════╣
║  🎯 Campaign-ready estimate: 1 046 / 1 247                   ║
╚══════════════════════════════════════════════════════════════╝

[After DQE server cleaning]

✅ Ready for campaign: 1 158 contacts

Remote source mode requires the dqe-one MCP server and at least one credential configured in DQEOne. Use --no-server or omit the MCP config to run local audit only.

---

dqe-list — Workspace overview

/dqe-quality:dqe-list

Calls the dqe-one MCP server and displays a structured overview of your DQEOne workspace:

• Processes — ID, name, type, draft status, last run
• Recent runs (last 5) — status, record count, errors, date
• Files — uploaded datasets with row/column counts
• Rulesets — available deduplication rulesets

Ends with a one-line workspace summary and suggested next steps.

Requires: dqe-one MCP server configured. See MCP setup.

---

dqe-dedup — Create a deduplication process

/dqe-quality:dqe-dedup [--source=csv|salesforce|dynamics|postgres|bigquery|sftp]

Guides you through creating a check_duplicate process on the DQE One Server:

1. Loads available files, rulesets, and credentials in parallel
2. Asks for the data source (CSV, Salesforce, Dynamics, PostgreSQL, BigQuery, SFTP)
3. Asks for ruleset, primary key, and field mapping — one section at a time
4. Shows a full configuration summary for confirmation
5. Calls create_process and optionally triggers the first run

# Start with a specific source pre-selected
/dqe-quality:dqe-dedup --source=csv
/dqe-quality:dqe-dedup --source=salesforce

Requires: dqe-one MCP server configured.

---

dqe-run — Run a process

/dqe-quality:dqe-run [process_name_or_id]

Triggers execution of a DQE process and shows its status.

1. Lists processes (or matches the one you named)
2. Confirms the process details (warns if still in draft)
3. Calls run_process
4. Shows the run status immediately and interprets completed, running, or failed

# Pick from the list
/dqe-quality:dqe-run

# Run a specific process by name
/dqe-quality:dqe-run "Monthly dedup — CRM"

Requires: dqe-one MCP server configured.

---

DQE One Server — MCP setup

The three skills dqe-list, dqe-dedup, dqe-run, and the server-cleaning phase of dqe-campaign all require a DQE One Server instance configured as an MCP server named dqe-one.

Option A — Using the dqe-one-mcp CLI (recommended)

pip install dqe-one-mcp
dqe-one-mcp install https://your-dqe-instance.com

This writes the MCP config to .claude/settings.json automatically.

Option B — Manual configuration

Add this to your .claude/settings.json:

{
  "mcpServers": {
    "dqe-one": {
      "type": "http",
      "url": "https://your-dqe-instance.com/mcp",
      "headers": {
        "Authorization": "Basic <base64(email:api_token)>"
      }
    }
  }
}

Replace <base64(email:api_token)> with the output of:

echo -n "your@email.com:your_api_token" | base64

Get your API token from My Profile in the DQE One web interface, or from the Setup page at /mcp-setup/.

---

The 6 dimensions

#	Dimension	What it detects
1	Completeness	Fill rate per column, globally empty fields
2	Invalid dates	Format errors, future dates, impossible values, mixed formats
3	Duplicates	Exact duplicates, near-duplicates (name+email, name+address)
4	Anomalies	Statistical outliers, generic values (null, test, xxx…), digits in text fields
5	Broken relationships	Postal code format per country (FR/DE/ES/US), ZIP/city mismatches, unreachable contacts
6	Format inconsistencies	Mixed phone formats, inconsistent casing, type heterogeneity

---

The reports

📊 Audit Report (dqe-audit — always generated)

• Quality score (0–100) with colour-coded rating
• Column profiling: detected type, fill rate, dominant value type
• Executive summary with 6 dimension cards
• Detailed per-dimension analysis with complete anomaly tables
• DQE service recommendations based on findings
• Next Steps — numbered action list from actual findings
• Contact DQE Software CTA block

⚙️ Project Manager Guide (dqe-audit — with --pm)

Advanced technical document for DQE internal teams:

• Detected parameters: encoding, delimiter, row/column counts
• Full column schema with top values and type distribution
• Prioritised treatment plan per dimension
• Technical configuration cards per relevant DQE service

---

Options

--lang — Report language (dqe-audit, dqe-campaign)

Value	Language
en	English (default)
us	English (alias)
fr	French
de	German
es	Spanish

--pm — Project Manager guide (dqe-audit only)

Generates the internal PM guide alongside the audit report.

--no-server — Local audit only (dqe-campaign only)

Skips all MCP server calls. Runs the local analysis only and stops after the audit card.

--source=<type> — Pre-select data source (dqe-campaign, dqe-dedup)

Skips the source selection question. Accepted values: csv, salesforce, dynamics, postgres, bigquery, snowflake, sftp.

---

Output files

Generated by dqe-audit next to the source CSV:

<basename>_dqe_audit_YYYYMMDD_<lang>.html        # always
<basename>_dqe_pm_guide_YYYYMMDD_<lang>.html     # with --pm

Collision guard: if a file with the same name already exists, a numeric suffix is added automatically.

---

File size handling

File size	Behaviour
Up to 200k rows	Full analysis — every row
200k – 500k rows	Auto-sampled (1 in N rows), result extrapolated
Above 500k rows	Rejected — tip provided to extract a sample

---

Requirements

• Claude Code CLI, desktop app, or IDE extension
• Python 3.x — standard library only, no pip install needed
• For MCP skills: a running DQE One Server instance

---

License

MIT — © 2026 DQE Software