pre-mortem

# Pre-Mortem Skill

**YOU MUST EXECUTE THIS WORKFLOW. Do not just describe it.**

Simulate implementation failures BEFORE building to catch problems early.

## Execution Steps

Given `/pre-mortem <spec-or-plan>`:

### Step 1: Load the Spec/Plan

If a path is provided, read it:
```
Tool: Read
Parameters:
  file_path: <provided path>
```

If no path, check for recent plans:
```bash
ls -lt .agents/plans/ .agents/specs/ 2>/dev/null | head -5
```

### Step 2: Understand What We're Building

Read the spec/plan carefully. Identify:
- **Goal**: What are we trying to build?
- **Components**: What pieces need to be created?
- **Integrations**: What does this connect to?
- **Constraints**: What limitations exist?

### Step 3: Dispatch Failure Expert Swarm

**Launch parallel failure simulations from different perspectives.**

```
Tool: Task
Parameters:
  subagent_type: "agentops:integration-failure-expert"
  description: "Integration failure simulation"
  prompt: |
    Simulate integration failures for this spec/plan:
    <spec-content>

    Identify: API mismatches, protocol issues, system boundary problems.
    Return findings with severity ratings.
```

```
Tool: Task
Parameters:
  subagent_type: "agentops:ops-failure-expert"
  description: "Operations failure simulation"
  prompt: |
    Simulate production operations failures for this spec/plan:
    <spec-content>

    Identify: Deployment risks, scaling issues, monitoring gaps, incident response holes.
    Return findings with severity ratings.
```

```
Tool: Task
Parameters:
  subagent_type: "agentops:data-failure-expert"
  description: "Data failure simulation"
  prompt: |
    Simulate data failures for this spec/plan:
    <spec-content>

    Identify: Corruption risks, consistency issues, migration problems, state management gaps.
    Return findings with severity ratings.
```

```
Tool: Task
Parameters:
  subagent_type: "agentops:edge-case-hunter"
  description: "Edge case hunting"
  prompt: |
    Hunt for unhandled ed