SkillAgentSearch skills...

chaos-engineering

Failure injection patterns, blast radius control, steady state hypothesis, and gameday planning for resilience testing.

Install / Use

npx skills add vibeeval/vibecosystem --skill chaos-engineering

Installs into whichever agent you are using.

About this skill
📄

SKILL.md

Installable skill definition

Quality Score

83/100

Category

Other

Supported Platforms

Universal

Our assessment of chaos-engineering

chaos-engineering scores 83/100 on our quality scale, 267th of 330 Other skills we index.

Its SKILL.md is 7.1 KB long, well organised into 16 sections with 5 code examples: a thorough specification that gives an agent plenty to work with.

It has 535 GitHub stars, a meaningful sign that others use it.

Substance
29/30
Structure
20/20
Description
12/15
Adoption
12/20
Freshness
11/15

Maintenance, license and trust

  • The repository was last updated about 3 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.
  • Our last check on 2026-11-07 found the source still online.
  • It is released under the MIT license, a permissive license that allows use, modification and commercial use with attribution.
  • Its trust signals score 98/100, with no cautions. These come from repository metadata, not a code audit — read the skill file before letting an agent act on it.

Safety scan

Review

Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review judged it risky: The skill instructs the agent to execute destructive infrastructure commands such as `kubectl delete pod`, `failover_db_cluster`, `fallocate` disk fill, `iptables` network partitions, and `timedatectl` clock skew.

AI review: risky

  • The skill instructs the agent to execute destructive infrastructure commands such as `kubectl delete pod`, `failover_db_cluster`, `fallocate` disk fill, `iptables` network partitions, and `timedatectl` clock skew.
  • As a skill/rules file, these are commands the agent would follow, so they create meaningful risk of service disruption or data issues if run in the wrong environment or without proper safeguards.
  • It is not malicious because it does not steal data, hide actions, or subvert instructions; it is a legitimate chaos-engineering guide with blast-radius controls and safety checklists.

AI review by kimi-k2.7-code on 2026-10-15. Automated pattern scan on 2026-10-15. It catches known dangerous patterns, not every risk — read a skill before letting an agent act on it.

chaos-engineering compared with similar skills

All 4 of these similar skills score higher than chaos-engineering; compare them before choosing.

SkillScoreStarsUpdatedFormat
chaos-engineering (this skill)by vibeeval835353mo agoSKILL.md
Agent-Reachby Panniantong100112.8k10d agoCLAUDE.md
headroomby headroomlabs-ai10079.9ktodayCLAUDE.md
Scraplingby D4Vinci100104.0ktodayMCP Server
crawl4aiby unclecode10087.5ktodayMCP Server

Frequently asked questions

How do I install chaos-engineering?
Run npx skills add vibeeval/vibecosystem --skill chaos-engineering. The install tabs above show the steps for each supported agent.
Which AI agents does chaos-engineering work with?
It is written for Universal, as a SKILL.md file. Other agents that read the same format can often use it too.
Is chaos-engineering safe to use?
Our scan of the whole file found no instruction hijacking, hidden characters, credential access, data exfiltration or destructive commands. An AI review judged it risky: The skill instructs the agent to execute destructive infrastructure commands such as kubectl delete pod, failover_db_cluster, fallocate disk fill, iptables network partitions, and timedatectl clock skew. It is MIT-licensed and scores 98/100 on trust signals. Skills are instructions an agent will follow, so read the file before installing it and do not approve commands you do not understand.
Is chaos-engineering still maintained?
The repository was last updated about 3 months ago. That is recent enough to be usable, but agent tooling moves fast, so check the instructions against your agent's current version.

name: chaos-engineering description: Failure injection patterns, blast radius control, steady state hypothesis, and gameday planning for resilience testing.

Chaos Engineering

Systematic resilience testing to discover weaknesses before they cause outages.

Steady State Hypothesis

# Define BEFORE injecting chaos - what "normal" looks like
steady_state_hypothesis:
  title: "API serves traffic within SLO"
  probes:
    - name: "API response time p95 < 500ms"
      type: http
      url: "https://api.example.com/health"
      threshold: 500

    - name: "Error rate < 1%"
      type: prometheus
      query: "rate(http_requests_total{status=~'5..'}[5m]) / rate(http_requests_total[5m])"
      threshold: 0.01

    - name: "Order processing queue depth < 100"
      type: cloudwatch
      metric: "ApproximateNumberOfMessagesVisible"
      threshold: 100

    - name: "Database connections < 80% capacity"
      type: prometheus
      query: "pg_stat_activity_count / pg_settings_max_connections"
      threshold: 0.8

Failure Injection Patterns

# Using Chaos Toolkit (chaostoolkit.org)
# experiment.json

{
  "title": "Database failover resilience",
  "description": "Verify app handles primary DB failover gracefully",

  "steady-state-hypothesis": {
    "title": "API responds normally",
    "probes": [
      {
        "name": "api-health",
        "type": "probe",
        "provider": {
          "type": "http",
          "url": "https://api.example.com/health",
          "timeout": 5
        },
        "tolerance": {"status": 200}
      }
    ]
  },

  "method": [
    {
      "name": "failover-primary-db",
      "type": "action",
      "provider": {
        "type": "python",
        "module": "chaosaws.rds.actions",
        "func": "failover_db_cluster",
        "arguments": {
          "db_cluster_identifier": "prod-cluster"
        }
      },
      "pauses": {"after": 60}
    }
  ],

  "rollbacks": [
    {
      "name": "verify-db-recovered",
      "type": "probe",
      "provider": {
        "type": "python",
        "module": "chaosaws.rds.probes",
        "func": "cluster_status",
        "arguments": {
          "db_cluster_identifier": "prod-cluster"
        }
      },
      "tolerance": "available"
    }
  ]
}

Blast Radius Control

# ALWAYS limit the impact of chaos experiments

class BlastRadiusController:
    """Control and limit chaos experiment impact."""

    def __init__(self, config: dict):
        self.max_affected_percentage = config.get('max_affected_pct', 5)
        self.max_duration_seconds = config.get('max_duration_s', 300)
        self.excluded_services = config.get('excluded', ['auth', 'payments'])
        self.kill_switch_url = config.get('kill_switch_url')

    def can_inject(self, target: str, scope: str) -> bool:
        # Never chaos-test critical services without explicit approval
        if target in self.excluded_services:
            return False

        # Never inject during peak hours
        hour = datetime.now().hour
        if 9 <= hour <= 17:  # Business hours (adjust per timezone)
            return False

        # Never affect more than N% of instances
        if self.get_affected_percentage(target, scope) > self.max_affected_percentage:
            return False

        return True

    def get_affected_percentage(self, target: str, scope: str) -> float:
        total = self.get_total_instances(target)
        affected = self.get_affected_instances(target, scope)
        return (affected / total) * 100 if total > 0 else 100

    async def emergency_stop(self) -> None:
        """Kill switch: immediately halt all chaos experiments."""
        await httpx.post(self.kill_switch_url, json={"action": "stop_all"})

Common Chaos Experiments

# Experiment catalog - start with these

level_1_basic:
  - name: "Kill a single pod"
    tool: "kubectl delete pod <name>"
    validates: "Pod auto-recovery, health checks"
    blast_radius: "1 pod"

  - name: "CPU stress on one node"
    tool: "stress-ng --cpu 4 --timeout 60"
    validates: "Autoscaling, request routing"
    blast_radius: "1 node"

  - name: "Inject 500ms network latency"
    tool: "tc qdisc add dev eth0 root netem delay 500ms"
    validates: "Timeout handling, circuit breakers"
    blast_radius: "1 container"

level_2_intermediate:
  - name: "Kill entire availability zone"
    tool: "Chaos Toolkit / AWS FIS"
    validates: "Multi-AZ failover, data replication"
    blast_radius: "1 AZ"

  - name: "DNS resolution failure"
    tool: "iptables -A OUTPUT -p udp --dport 53 -j DROP"
    validates: "DNS caching, fallback resolution"
    blast_radius: "1 service"

  - name: "Disk fill to 95%"
    tool: "fallocate -l 50G /tmp/disk_fill"
    validates: "Disk space alerts, log rotation"
    blast_radius: "1 node"

level_3_advanced:
  - name: "Split brain network partition"
    tool: "Toxiproxy / Linux iptables"
    validates: "Consensus protocols, data consistency"
    blast_radius: "Cluster segment"

  - name: "Clock skew injection"
    tool: "timedatectl set-time +5min"
    validates: "Certificate validation, token expiry"
    blast_radius: "1 node"

Gameday Checklist

## Pre-Gameday (1 week before)
- [ ] Define steady state hypothesis with measurable probes
- [ ] Identify blast radius and set hard limits
- [ ] Ensure kill switch is tested and accessible
- [ ] Notify on-call team and stakeholders
- [ ] Verify rollback procedures are documented and tested
- [ ] Set up monitoring dashboards for the experiment
- [ ] Run experiment in staging first

## During Gameday
- [ ] Verify steady state BEFORE injecting chaos
- [ ] Start with smallest blast radius, escalate gradually
- [ ] Monitor dashboards continuously during experiment
- [ ] Document observations in real-time (shared doc)
- [ ] If SLO violated: trigger kill switch immediately
- [ ] Time-box each experiment (max 5 minutes per injection)

## Post-Gameday
- [ ] Verify system returned to steady state
- [ ] Document findings: what broke, what recovered, what surprised
- [ ] Create action items for discovered weaknesses
- [ ] Update runbooks based on learnings
- [ ] Share results with broader engineering team
- [ ] Schedule fixes and re-test

Checklist

  • [ ] Define steady state hypothesis before every experiment
  • [ ] Never run chaos in production without a tested kill switch
  • [ ] Start in staging, graduate to production with reduced blast radius
  • [ ] Exclude critical services (auth, payments) unless specifically targeting them
  • [ ] Time-box experiments (max 5 minutes injection, 30 minutes observation)
  • [ ] Run during low-traffic windows, never during peak
  • [ ] Document every experiment: hypothesis, method, observations, findings
  • [ ] Automate recurring experiments in CI/CD pipeline

Anti-Patterns

  • Chaos without hypothesis: random breaking is not engineering
  • No kill switch: unable to stop experiment when things go wrong
  • Running in production first: always validate in staging
  • Affecting too many instances: never exceed 5% without explicit approval
  • Chaos during incidents: only inject chaos on healthy systems
  • Not fixing findings: experiments without follow-up action items are wasted

Related Skills

View on GitHub
GitHub Stars535
CategoryOther
Updated3mo ago
Forks44

Languages

C#

Trust signals

98/100

From repository metadata: license, adoption, age and documentation. Not a code audit — see the Safety scan above for what the skill file itself contains.

1 info