Skip to content
Get packs

Guide · Agent skill

How to use the Managed Airflow Recommendations agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

This guide explains how to install and use the managed-airflow-recommendations agent skill, a set of instructions and bundled Python scripts that let an AI agent review a Google Cloud Managed Service for Apache Airflow (formerly Cloud Composer) environment. It is written for readers who have not used agent skills before. By the end you will know what the skill does, how to install it, how to ask for a review in plain language, and what you need in place first.

Managed Airflow Recommendationsgoogle/skills · Apache-2.0

Quick start

  1. 1Download and unzip the skill. Download the ZIP file from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
  2. 2Install and authenticate gcloud. Install the Google Cloud SDK, then run gcloud auth login and gcloud auth application-default login.
  3. 3Enable the required APIs. Enable Composer, Cloud Monitoring and Cloud Logging on the project that hosts your environment.
  4. 4Confirm your IAM roles. Make sure you have the read-only Composer, monitoring viewer and logging viewer roles on the project.
  5. 5Gather your environment details. Note the environment name, project ID and region, for example us-central1.
  6. 6Ask for a review. Describe the goal in plain language, such as a sizing check or a restart investigation, and include the environment details.
  7. 7Act on the recommendations. Read the categorized recommendations, check the data the agent cites, and apply changes after you have reviewed them.

What is an agent skill, and what does this one do?

An agent skill is a folder of instructions and supporting files that an AI agent can load when a task matches the skill's description. This one, named managed-airflow-recommendations, covers recommendations and best practices for creating, configuring, tuning and optimizing Managed Airflow environments.

When installed, the agent reads the environment's configuration with gcloud, runs bundled Python scripts that pull CPU, memory, disk, restart and DAG parsing metrics from Cloud Monitoring, and then suggests scaling and tuning changes based on that data rather than guessing. The skill is designed to avoid recommending a bigger size by default.

  • Reviews sizing, scheduler and worker stability, and DAG parsing efficiency.
  • Gathers evidence from Cloud Monitoring before recommending changes.
  • Does not troubleshoot a single failed DAG or task run, write or migrate DAG code.

Install the skill from the ZIP file

Download the skill's ZIP file from the page where you found this guide. Unzip it into one of two locations, depending on how you want to use it. Put it in ~/.claude/skills/ to make it available across your projects, or in a project's .claude/skills/ folder to limit it to that project.

If you use the Claude apps instead, you can upload the skill there under Settings → Capabilities → Skills.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into the project's .claude/skills/ folder
  • Claude apps: Settings → Capabilities → Skills, then upload the skill

Set up Google Cloud before you start

The scripts and gcloud commands fail without these prerequisites, so check them first. You need the Google Cloud SDK (gcloud) installed and authenticated, including application default credentials. The scripts need Python 3.9 or later and use only the standard library, so there are no extra packages to install.

You also need the Managed Airflow, Cloud Monitoring and Cloud Logging APIs enabled on the project, and read-only roles: Composer environment and storage object viewer, monitoring viewer, and logging viewer. Have your environment's project ID and region ready.

  • Google Cloud SDK (gcloud) installed and authenticated
  • Application default credentials set up
  • Python 3.9 or later
  • APIs enabled: Composer, Cloud Monitoring, Cloud Logging
  • Read-only IAM roles: Composer environment and storage object viewer, monitoring viewer, logging viewer
  • An existing Managed Airflow environment with its project ID and region
gcloud auth login gcloud auth application-default login gcloud config set project {env_project} gcloud services enable composer.googleapis.com monitoring.googleapis.com logging.googleapis.com

Ask for a review with a natural request

You do not need special syntax. Describe what you want in plain words and include the environment name, project ID and region. If you leave any of these out, the skill asks you for them. If you are looking into a recent problem, say over what period, such as the past 24 hours or the past 7 days.

Useful request types include a sizing check, a stability review after restarts, a memory pressure question, a DAG parsing question, or a general health check. The agent will then gather data before it makes any recommendations.

  • "Review the sizing of my Managed Airflow environment my-env in project my-project, region us-central1."
  • "Why is my scheduler restarting? Check the past 7 days."
  • "Do a health check and best-practice review of my Composer environment."
  • "Is the DAG parsing time in my-env too high?"

What the agent does during a review

The skill sets out a fixed workflow. The agent first confirms the target environment and timeframe, then retrieves the environment configuration, such as its size, number of schedulers, max workers and CPU and RAM limits. It then runs the scripts in the scripts/ directory to gather telemetry and does not make generalizations without data.

After analysis, the agent presents a categorized list of recommendations. The categories are Infrastructure & Scaling, Airflow Configurations, Bucket Hygiene, and Production Best Practices. Recommendations to consider include workload count, resources, parallelism settings and removing non-DAG files from the dags/ folder.

  • Infrastructure & Scaling
  • Airflow Configurations
  • Bucket Hygiene
  • Production Best Practices
python3 scripts/environment_health.py --help

Safety notes: scripts and network use

This skill runs scripts and uses the network. The bundled scripts call Cloud Monitoring and Cloud Logging, and the gcloud commands read your environment's configuration. Use it only on environments you are allowed to read. The scripts need monitoring and configuration permissions and will fail without them.

The skill also asks the agent to tag its gcloud commands with an attribution variable that identifies only the skill and its version. The tag is set inline on each command rather than saved with gcloud config set. The lib/ folder contains shared helpers and should not be run directly. Review the scripts before installing if your organisation requires it.

  • Read-only roles are enough; the skill is built around least privilege.
  • Do not run the files in scripts/lib/ directly.
  • The skill does not troubleshoot single DAG or task failures and does not write DAG code.

What people use it for

  • Checking whether workers are consistently maxing out CPU and need more workload count or resources.
  • Investigating scheduler or worker restarts and whether memory pressure is the cause.
  • Reviewing DAG parsing times to see whether the scheduler is slowed by heavy top-level code.
  • Running a general health check and best-practice review before a production launch.
  • Finding non-DAG files in the dags/ folder that should be removed.

Frequently asked questions

Is the managed-airflow-recommendations skill free?

The skill is published under the Apache-2.0 licence. Google Cloud usage, including the Managed Airflow environment and Cloud Monitoring, is billed under your own Google Cloud account.

What do I need before I can use this skill?

You need the Google Cloud SDK installed and authenticated with application default credentials, Python 3.9 or later, the Composer, Cloud Monitoring and Cloud Logging APIs enabled, read-only IAM roles, and an existing environment with its project ID and region.

Where do I install the skill?

Unzip the downloaded ZIP into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps you can upload it under Settings → Capabilities → Skills.

Does this skill fix a failed DAG run?

No. The skill explicitly excludes troubleshooting individual DAG or task run failures. It focuses on environment sizing, stability, parsing efficiency and configuration.

Will the skill just tell me to make my environment bigger?

It is designed not to. The skill analyses CPU, memory, disk, restart and parsing metrics first and looks for tuning options before recommending a larger size.

How is this different from using gcloud alone?

gcloud shows configuration, but the skill adds a workflow that pulls metrics from Cloud Monitoring with bundled scripts and turns them into categorized recommendations. You still need gcloud installed and authenticated for it to work.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits data engineers and platform teams who run Managed Airflow (Cloud Composer) and want an evidence-based sizing, stability or health review using read-only access. It is not the right choice for debugging a single failed DAG, writing or migrating DAG code, or for environments you cannot read. For those tasks, use Airflow's own tooling or a DAG-focused resource instead.

Download the skill