Steadybit agent skills for Claude Code, Cursor and RevampKey

5 ready-made playbooks that teach an AI agent how to do real Steadybit work: which tools to call, what to check first, and when to stop and ask you.

Download RevampKey, free I have it: connect Steadybit

Connect Steadybit once and all 5 skills install themselves and stay up to date. Mac, Windows and Linux.

All 5 Steadybit skills

Each skill is one job. You ask in plain words and the agent loads the skill that matches, then follows its steps.

Start hereAsks before writing24 Steadybit tools

Use Steadybit from your AI assistant, safely

Connects an AI assistant to your Steadybit tenant so it can explore environments, services, experiments and execution results. It confirms the tenant and permission scope first and asks before running, cancelling or saving anything.

Use it when

You want to ask your AI assistant what Steadybit can see in your environments and teams
You are connecting Steadybit for the first time and want to know what the assistant can and cannot change
A Steadybit query fails or returns too much data and you need to narrow it

Tools it uses

list_teams
list_environments
get_target_stats
list_templates
get_template
get_action
get_service
get_service_risk_report
get_target
get_reliability_advice
get_experiment
get_execution
get_execution_log
get_execution_metric
list_execution_metrics
list_metric_label_values
get_execution_artifact
get_execution_report
analyze_execution
draft_experiment
create_experiment
create_experiment_from_template
run_experiment
cancel_execution

Example

Connect to Steadybit and show me which teams and environments I can see, then list the services in the production environment.

You can see 3 teams and 2 environments through your credential: Payments and Checkout in production, and Payments in staging. Production lists 14 services. Nothing has been changed; I will not run or save anything without your explicit yes.

Use in RevampKeyRead the full skill →

Asks before writing11 Steadybit tools

Review a Steadybit experiment execution with AI

Goes through a finished Steadybit execution step by step: which steps failed, which targets were hit, what the logs and metrics show, and whether the system passed. It only reads results and never starts or changes a run.

Use it when

An experiment execution failed and you need to know why
You want to know whether a chaos run actually passed
Metrics from a fault injection need interpreting

Tools it uses

get_execution
get_experiment
get_execution_report
get_execution_log
get_execution_artifact
list_execution_metrics
get_execution_metric
list_metric_label_values
get_reliability_advice
run_experiment
cancel_execution

Example

Why did execution 4821 of the checkout latency experiment fail?

The checkout step failed, not the fault injection. Of 10 planned targets, 8 were hit. The payment-service log shows its first timeout 14 seconds into the fault, and p95 latency went from 210 ms at baseline to 2.4 s during the window. It recovered within 40 seconds after the fault ended. The verdict is failed on resilience: payment-service has no retry on its upstream call. Two targets were not reached, so the scope needs checking as well.

Use in RevampKeyRead the full skill →

Asks before writing12 Steadybit tools

Design a Steadybit experiment from a resilience question

Turns a resilience question into a draft experiment for a chosen service, checks it against the platform, and saves it only after you approve the design. It does not run anything.

Use it when

A team asks what happens when a dependency fails and wants a repeatable test
You want to start from an existing template and adapt it to one service
A design needs a clear hypothesis, abort conditions and a small blast radius before anyone runs it

Tools it uses

list_environments
get_service
list_templates
get_template
get_target
get_action
get_experiment
draft_experiment
create_experiment
create_experiment_from_template
run_experiment
list_execution_metrics

Example

Design an experiment to check that checkout keeps working if the payment service is slow in staging.

Proposed hypothesis: when latency of 800 ms hits one payment-service instance in staging, checkout error rate stays below 2% for the 10-minute window. I found an existing template for service latency and adapted it to that one instance, with an abort if error rate exceeds 5%. Nothing has been saved yet. Shall I create it as a draft experiment? It will not run.

Use in RevampKeyRead the full skill →

Asks before writing11 Steadybit tools

Check whether a Steadybit experiment is safe to run

Reviews the target scope, environment, service state and existing executions before a fault injection starts, and asks for your explicit approval before running or cancelling anything.

Use it when

Before starting a chaos experiment in an environment you have not run before
When a service is already under pressure and you want to know whether a test is wise
When you need to confirm that an experiment only hits the targets you intended

Tools it uses

get_experiment
list_templates
list_environments
get_target
get_target_stats
get_service
cancel_execution
get_service_risk_report
get_execution
list_execution_metrics
run_experiment

Example

Is the checkout latency experiment ready to run in staging, or would it hit more pods than we meant?

The checkout latency experiment is in staging and targets 3 of 12 pods in the checkout deployment, which matches the intent you described. The service is healthy and there is no overlapping execution. There is no automatic stop condition, so the run can only be halted manually. Verdict: ready with one condition. Shall I start it once you confirm?

Use in RevampKeyRead the full skill →

Read-only9 Steadybit tools

Triage service risk in Steadybit with AI

Ranks your services by resilience risk, checks coverage and recent trends, and proposes the next test for each. It reads the risk picture and does not start any run without asking.

Use it when

You want to know which services are riskiest across your environments.
You need to decide where to focus resilience testing this sprint.
You want to see whether reliability is improving or getting worse over time.

Tools it uses

list_environments
list_teams
get_service
get_service_risk_report
get_reliability_advice
list_templates
get_experiment
list_execution_metrics
get_execution_metric

Example

Which of our services are the riskiest right now, and what should we test next in production?

Checkout-api ranks first: high risk driven by no timeout coverage on its payment dependency, with no experiments yet, so the first test is a slow-payments fault. Inventory ranks second: high risk and failing its last two runs, so fix those before adding new tests. Five other services are low risk or have no data, and I have marked them as not assessed.

Use in RevampKeyRead the full skill →

Get all 5 without copying a file

RevampKey adds the skills when you connect Steadybit, updates them for you and removes them when you disconnect.

Download RevampKey

How Steadybit skills work

Connect Steadybit
The Steadybit MCP server gives the agent its tools.
Ask in plain words
“Why did execution 4821 of the checkout latency experiment fail?” Each SKILL.md says when it applies, so only the one that fits is loaded.
Approve any change
The agent calls the tools in order, checks its work, and shows you any change before it makes it.

How to install Steadybit skills

In RevampKey the skills install themselves when you connect Steadybit. In other tools each skill is one folder with a SKILL.md file, and you also need the Steadybit MCP server connected there.

About 1 minute · nothing to copy

Download RevampKey for Mac, Windows or Linux and sign in.
Open Connectors and connect Steadybit. All 5 skills come with it and stay up to date; they go away when you remove the connector.
Ask in plain language, for example “Why did execution 4821 of the checkout latency experiment fail?”
Download RevampKeyConnect Steadybit · 1 click
Get RevampKey to connect Steadybit
RevampKey is a free desktop app for Mac, Windows and Linux. Install it, sign in, and Steadybit is one click away with its skills already attached.
Download RevampKey
Not now? Use the skills in your current tool
for s in essentials execution-results-review experiment-design-from-intent experiment-run-readiness service-risk-triage; do curl -fsSL --create-dirs -o .claude/skills/steadybit-$s/SKILL.md https://revampkey.com/marketplace/steadybit/skills/$s/SKILL.md done
Other tools

Into this project’s .claude/skills folder

Connect the Steadybit MCP server in Claude Code.

Run this in your project folder to add all 5 skills:

for s in essentials execution-results-review experiment-design-from-intent experiment-run-readiness service-risk-triage; do
  curl -fsSL --create-dirs -o .claude/skills/steadybit-$s/SKILL.md https://revampkey.com/marketplace/steadybit/skills/$s/SKILL.md
done

Ask in plain language, for example “Why did execution 4821 of the checkout latency experiment fail?”

Into the shared .agents/skills folder

Connect the Steadybit MCP server in Cursor.

Run this in your project folder to add all 5 skills:

for s in essentials execution-results-review experiment-design-from-intent experiment-run-readiness service-risk-triage; do
  curl -fsSL --create-dirs -o .agents/skills/steadybit-$s/SKILL.md https://revampkey.com/marketplace/steadybit/skills/$s/SKILL.md
done

Ask in plain language, for example “Why did execution 4821 of the checkout latency experiment fail?”

Into the shared .agents/skills folder

Connect the Steadybit MCP server in Codex.

Run this in your project folder to add all 5 skills:

for s in essentials execution-results-review experiment-design-from-intent experiment-run-readiness service-risk-triage; do
  curl -fsSL --create-dirs -o .agents/skills/steadybit-$s/SKILL.md https://revampkey.com/marketplace/steadybit/skills/$s/SKILL.md
done

Ask in plain language, for example “Why did execution 4821 of the checkout latency experiment fail?”

Into the shared .agents/skills folder

Connect the Steadybit MCP server in VS Code.

Run this in your project folder to add all 5 skills:

for s in essentials execution-results-review experiment-design-from-intent experiment-run-readiness service-risk-triage; do
  curl -fsSL --create-dirs -o .agents/skills/steadybit-$s/SKILL.md https://revampkey.com/marketplace/steadybit/skills/$s/SKILL.md
done

Ask in plain language, for example “Why did execution 4821 of the checkout latency experiment fail?”

RevampKey or copy the files?

In RevampKey

One click when you connect Steadybit
Updated for you when a skill improves
Removed when you disconnect
The MCP server is set up for you
800+ other connectors, same way

Download free

Copy into your project

One command per project
Run it again to get updates
You connect the Steadybit MCP server yourself
Works in Claude Code, Cursor, Codex and VS Code

See the command

Steadybit skills FAQ

What is a Steadybit agent skill?

A SKILL.md file: a short playbook an AI agent loads when your request matches it. Each Steadybit skill describes one job, which Steadybit tools to use in which order, what to check first and when to stop and ask you.

Do the Steadybit skills change anything on their own?

1 of the 5 only read. The others tell the agent to show you the change and wait for a yes before it creates, edits or deletes anything in Steadybit.

Which tools can use these skills?

Any agent that reads the open Agent Skills format: Claude Code, Cursor, Codex, VS Code with Copilot and RevampKey. In RevampKey they are attached when you connect Steadybit; elsewhere you copy the folders into your project or skills directory.

Do I need the Steadybit MCP server too?

Yes. The skills tell the agent how to do the work; the Steadybit MCP server gives it the tools to do it. The Steadybit MCP setup guide covers every client.

Steadybit MCP server setup · This page as Markdown

Put Steadybit’s best workflows in your agent

Download RevampKey, connect Steadybit and ask your first question in about a minute.

Download RevampKey Connect Steadybit · 1 click

All 5 Steadybit skills

Use Steadybit from your AI assistant, safely

Use it when

Tools it uses

Review a Steadybit experiment execution with AI

Use it when

Tools it uses

Design a Steadybit experiment from a resilience question

Use it when

Tools it uses

Check whether a Steadybit experiment is safe to run

Use it when

Tools it uses

Triage service risk in Steadybit with AI

Use it when

Tools it uses

Get all 5 without copying a file

How Steadybit skills work

How to install Steadybit skills

RevampKey or copy the files?

In RevampKey

Copy into your project

Steadybit skills FAQ

Put Steadybit’s best workflows in your agent

Skills for related tools