felipecustodio Overview & Purpose
Causal Analysis moves beyond superficial symptoms to isolate underlying root causes. It distinguishes between correlation and causation, helping agents prevent recurring failures by addressing foundational systemic issues.
When to Use
- Incident Post-Mortems: Diagnosing production outages or major system bugs.
- System Failure Diagnosis: Distinguishing root causes from surface-level errors.
- Policy & Process Improvements: Understanding downstream effects of structural changes.
Execution Workflow
- Symptom Mapping: State the observed failure clearly.
- The 5 Whys Traversal: Iteratively ask "Why did this occur?" down 5 levels of causality.
- Counterfactual Test: Verify causality by asking: "If cause X was absent, would outcome Y still have occurred?".
- Root Cause Identification: Isolate the foundational driver.
- Preventative Action: Define corrective controls targeted directly at the root cause.
Expected Output Contract
### Causal Analysis Summary
- **Observed Symptom**: [Surface error]
- **Causal Chain (5 Whys)**:
1. Why? -> [Direct Cause]
2. Why? -> [Sub Cause]
...
5. Why? -> [Root Cause]
- **Root Cause**: [Core Issue]
- **Preventative Action**: [Structural Fix]
Scripts
scripts/causal_analysis.py- Deterministic evaluation, state validation, and CLI tool for causal-analysis.