akomar-mozdata-marketplace

Mozdata Plugin for Claude Code

A Claude Code plugin for Mozilla data engineering: telemetry probe discovery, BigQuery query writing, and Airflow DAG failure debugging.

This plugin is in development. Feel free to use and report issues.

What It Does

This plugin helps you:

• Discover telemetry probes - Find Glean metrics across Mozilla products (Firefox Desktop, Android, etc.)
• Write BigQuery queries - Generate efficient queries for Mozilla telemetry data
• Navigate data resources - Access Glean Dictionary, ProbeInfo API, and BigQuery schemas
• Debug Airflow failures - Investigate DAG failures, fetch task logs, and identify root causes

Features

/mozdata:ask Slash Command

Activates Mozilla telemetry expertise in your current conversation. Use it to:

• Find probes for specific products or features
• Get guidance on which BigQuery tables to query
• Generate SQL queries following Mozilla best practices
• Understand metric metadata and collection details

Example usage:

/mozdata:ask How do I query Firefox Desktop DAU?
/mozdata:ask Find probes related to accessibility in Firefox
/mozdata:ask Write a query for the a11y_hcm_foreground metric

Skills (Auto-Activated)

The plugin includes Skills that Claude can automatically invoke when relevant:

For Data Scientists & Analysts

mozilla-probe-discovery - Activated when you ask about:

• Finding metrics or probes in Mozilla products
• Glean telemetry data
• Accessibility probes, search metrics, etc.

mozilla-query-writing - Activated when you ask about:

• Firefox DAU/MAU queries
• BigQuery Mozilla telemetry
• baseline_clients_*, events_stream tables
• User counts or data analysis

For Data Engineers

airflow-debugging - Activated when you ask about:

• Failed DAGs or DAG run errors
• Airflow task logs and debugging
• bqetl pipeline failures
• Data pipeline debugging

This skill includes bundled scripts for listing failed DAGs and fetching task logs from GCS. It requires gcloud authentication and access to gs://airflow-remote-logs-prod-prod. It only activates when relevant — users who only need telemetry/query help won't encounter any additional prerequisites.

Skills activate automatically based on your questions - no need to use the slash command.

Architecture

mozdata-claude-plugin/
├── knowledge/                 # Shared knowledge modules
│   ├── architecture.md        # Data platform architecture
│   ├── metrics.md             # Metric/probe discovery
│   ├── data-catalog.md        # Table discovery
│   └── query-writing.md       # Query best practices
├── commands/
│   └── ask.md                 # Slash command (refs @knowledge/*)
├── skills/
│   ├── probe-discovery/
│   │   └── SKILL.md           # Auto-activated for probe questions
│   ├── query-writing/
│   │   └── SKILL.md           # Auto-activated for query questions
│   └── airflow-debugging/
│       ├── SKILL.md           # Auto-activated for Airflow failures
│       └── scripts/           # Helper scripts (list-failed-dags, fetch-task-log)
├── bigquery/
│   └── tools.yaml             # genai-toolbox config (BigQuery access)
├── .claude-plugin/
│   └── plugin.json
├── .mcp.json                  # Bundled MCP servers (Glean Dictionary)
└── README.md

Knowledge Modules

The plugin's intelligence is organized into four shared knowledge modules:

Module	Purpose
architecture.md	Data platform structure, Glean schema, gotchas
metrics.md	ProbeInfo API, Glean Dictionary, metric types
data-catalog.md	Table selection, aggregation hierarchy
query-writing.md	Required filters, templates, anti-patterns

Both the slash command and Skills reference these modules for consistent knowledge.

Key Knowledge

The plugin understands:

• Glean Dictionary structure and navigation
• ProbeInfo API endpoints and data formats
• BigQuery conventions (mozdata dataset, submission_date filtering, sample_id usage)
• Query best practices (aggregate tables, required filters, performance optimization)
• Mozilla data platform architecture and resources

Resources Used

• Glean Dictionary - Probe exploration UI
• ProbeInfo API - Programmatic probe metadata
• Mozilla Data Docs - Data platform documentation
• bigquery-etl - Query definitions and aggregates
• Glean Dictionary MCP - Direct access to Glean metadata (bundled with plugin)
• DataHub MCP - BigQuery schema metadata (requires separate setup)

Setup

Installation

Add the GitHub repository as a marketplace and install the plugin:

claude plugin marketplace add akkomar/mozdata-claude-plugin
claude plugin install mozdata@akomar-mozdata-marketplace

Prerequisites

DataHub MCP Server (for BigQuery schema access)