StartChaos — Orchestrated Chaos Experiment Pipeline

⛔ ABSOLUTE RULE: Do NOT improvise, skip, or substitute any step. On ANY error, STOP and wait for the user.

When to use this skill (vs. the MCP server)

This skill is the human-interactive path: it persists state, renders cards, prompts for missing inputs, and generates an HTML report at the end. Use it when there is a user in the loop.

If you are an autonomous agent with no user to prompt — or you need typed, atomic operations to compose into your own workflow — use the chaos-studio MCP server instead (see mcp/README.md). The MCP exposes the same Chaos Studio v2 surface as discrete tools (chaos_create_workspace, chaos_list_recommended_scenarios, chaos_execute_scenario, …) without the state file, cards, or interactive prompts.

Both surfaces target Microsoft.Chaos 2026-05-01-preview and use the local az login session for auth.

How It Works

All pipeline logic lives in scripts/Invoke-StartChaos.ps1. The script handles auth, workspace creation, scenario setup, execution, permission validation/fix, and the final summary card.

The AI orchestrator's only job is:

1. Set $env:STARTCHAOS_STATE_PATH to ${SESSION_DIR}/startchaos-state.json
2. Run the script
3. Handle exit codes that require user input (see below)
4. Re-run the script with the user's answers as parameters

Running the Script

$env:STARTCHAOS_STATE_PATH = "<session-dir>/startchaos-state.json"
& "<skill-dir>/scripts/Invoke-StartChaos.ps1" @args

On resume: the script reads state and skips completed phases automatically.

Exit Codes → AI Actions

Important: Exit 2 → 3 chaining

When exit 2 is followed by exit 3 on re-run, combine both user answers in the next invocation:

& Invoke-StartChaos.ps1 -ScenarioName "ZoneDown-1.0" -ParameterMode "autofill" -ParameterValues @{ duration = 'PT5M' }

Important: Manual mode does NOT work

-ParameterMode manual uses Read-Host prompts that do not render in a non-interactive AI session. Always use -ParameterMode autofill combined with -ParameterValues for any overrides the user requests. When exit 3 fires, show the user the parameter table from the output, ask which values they want to change, then re-run with autofill + overrides.

Script Parameters (passed through to sub-phases)

What the Script Handles (no AI logic needed)

• Phase sequencing and state-based resume
• Auth pre-flight (existing session detection, device-code login)
• Workspace creation, identity binding, RBAC grants
• Scenario evaluation, recommendation discovery, configuration creation
• Pre-run validation and fixResourcePermissions auto-remediation
• Run execution, polling, action status tracking
• Final summary card and HTML report generation
• All error cards with remediation commands