Chaos experiment
A collection of structured AI agent skills that enable Claude Code, Cursor, GitHub Copilot, and other AI coding assistants to create, operate, debug, and govern Harness CI/CD workflows through natural language.
npx -y skills add harness/harness-skills --skill chaos-experimentAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
What its author says it does
Copied from the file, not written here
Create and manage chaos experiments using Harness Chaos Engineering via MCP. Run resilience tests like pod deletion, CPU stress, and network faults. Use when user says "chaos experiment", "chaos engineering", "resilience test", "chaos test", or wants to test system reliability.
The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.
SKILL.md
3.7 KB, as published. Nobody here has run it
Chaos Experiment
Create and manage chaos experiments using Harness Chaos Engineering via MCP.
Instructions
Step 1: Check Infrastructure
Call MCP tool: harness_list
Parameters:
resource_type: "chaos_infrastructure"
org_id: "<organization>"
project_id: "<project>"
Step 2: Browse Templates
Call MCP tool: harness_list
Parameters:
resource_type: "chaos_experiment_template"
org_id: "<organization>"
project_id: "<project>"
Step 3: List Existing Experiments
Call MCP tool: harness_list
Parameters:
resource_type: "chaos_experiment"
org_id: "<organization>"
project_id: "<project>"
Step 4: Create Experiment
Call MCP tool: harness_create
Parameters:
resource_type: "chaos_experiment"
org_id: "<organization>"
project_id: "<project>"
body: <experiment definition>
Step 5: Run Experiment
Call MCP tool: harness_execute
Parameters:
resource_type: "chaos_experiment"
action: "run"
resource_id: "<experiment_id>"
org_id: "<organization>"
project_id: "<project>"
Step 6: Monitor Results
Call MCP tool: harness_list
Parameters:
resource_type: "chaos_experiment_run"
org_id: "<organization>"
project_id: "<project>"
Get specific run details:
Call MCP tool: harness_get
Parameters:
resource_type: "chaos_experiment_run"
resource_id: "<run_id>"
Step 7: Check Probes
Call MCP tool: harness_list
Parameters:
resource_type: "chaos_probe"
org_id: "<organization>"
project_id: "<project>"
Common Experiment Types
- Pod Delete - Kill pods to test recovery
- Pod CPU Hog - Stress CPU to test throttling
- Pod Memory Hog - Consume memory to test OOM handling
- Pod Network Loss - Simulate network failures
- Pod Network Latency - Add artificial latency
- Node Drain - Drain K8s nodes
- EC2 Stop - Stop AWS EC2 instances
- ECS Task Stop - Stop ECS tasks
Chaos Resource Types
| Resource Type | Operations | Description |
|---|---|---|
chaos_experiment | list, get, create, update, delete, run | Experiments |
chaos_experiment_run | list, get | Run history/results |
chaos_experiment_template | list, get | Pre-built templates |
chaos_infrastructure | list, get | Target infrastructure |
chaos_probe | list, get | Health probes |
Examples
- "Show me all chaos experiments" - List chaos_experiment
- "Create a pod-delete experiment for checkout-service" - Create chaos_experiment
- "Run the weekly resilience test" - Execute run action
- "What were the results of the last chaos run?" - Get chaos_experiment_run
Performance Notes
- Review existing experiments before creating duplicates. Check for similar fault types targeting the same service.
- Wait for experiment completion before analyzing results. Do not draw conclusions from partial runs.
- Verify the target infrastructure and service are healthy before running chaos experiments.
Troubleshooting
Experiment Won't Run
- Verify chaos infrastructure is connected and active
- Check target application/namespace exists
- Ensure RBAC permissions for chaos operations
Probes Failing
- Check probe endpoints are accessible
- Verify probe timeout settings
- Review probe type matches expected behavior