proffesor-for-testing/agentic-qe/exploratory-testing-advanced

Advanced Exploratory Testing

<default_to_action> When exploring software or investigating quality risks:

1. CREATE charter with mission, scope, and time-box (45-90 min)
2. APPLY heuristics: SFDIPOT (quality criteria), FEW HICCUPPS (consistency oracles)
3. EXPLORE systematically using test tours (Business District, Bad Neighborhood, Historical)
4. DOCUMENT findings in real-time with notes, screenshots, evidence
5. DEBRIEF: What learned? What's next? Share via agent memory

Quick Heuristic Selection:

• What to test → SFDIPOT (Structure, Function, Data, Interfaces, Platform, Operations, Time)
• Recognize problems → FEW HICCUPPS (Familiar, Explainable, World, History, Image, Comparable, Claims, Users, Product, Purpose, Standards)
• Navigate app → Test Tours (12 types for different exploration strategies)

Critical Success Factors:

• Exploration is skilled, structured thinking - not random clicking
• Document discoveries, not pre-planned test cases
• Pair testing reveals more than solo exploration

</default_to_action>

Quick Reference Card

When to Use

• Investigating new or changed features
• Finding bugs automation misses
• Learning unfamiliar systems
• Risk discovery before test planning

Session Structure (SBTM)

SFDIPOT Heuristic (What to Test)

FEW HICCUPPS Oracle (Recognize Problems)

Test Tours (12 Types)

Session Note Template

**Charter:** Explore [area] to discover [what] focusing on [heuristic]
**Time-box:** 60 min | **Tester:** [name] | **Date:** [date]
 
## Session Notes
-[timestamp] Observation/finding
-[timestamp] Bug: [description] - [severity]
-[timestamp] Question: [unclear behavior]
 
## Findings Summary
-Bugs: X (Critical: Y, Major: Z)
-Questions: X
-Ideas: X
 
## Coverage
-Areas explored: [list]
-Heuristics used: [SFDIPOT areas]
-% Time on: Bug investigation 30%, Exploration 50%, Setup 20%
 
## Next Steps
-[ ] Deep dive on [area]
-[ ] Follow up on question about [topic]

Agent-Assisted Exploration

// Collaborative exploration session
await Task("Exploratory Session", {
  charter: 'Explore checkout flow for payment edge cases',
  duration: '60min',
  heuristics: ['SFDIPOT', 'FEW_HICCUPPS'],
  tour: 'money',
  collaboration: 'human-navigator-agent-driver'
}, "qe-flaky-test-hunter");
// Agent generates test variations while human observes
await Task("Edge Case Generation", {
  area: 'payment-form',
  variations: ['boundary-values', 'invalid-inputs', 'concurrent-submits']
}, "qe-test-generator");
// Visual exploration
await Task("Visual Exploration", {
  tour: 'landmark',
  focus: 'responsive-breakpoints',
  compare: 'baseline-screenshots'
}, "qe-visual-tester");

Agent Coordination Hints

Memory Namespace

aqe/exploratory/
├── sessions/*           - Session notes and findings
├── charters/*           - Reusable charter templates
├── bug-clusters/*       - Historical bug patterns
└── heuristic-results/*  - What heuristics revealed

Fleet Coordination

const exploratoryFleet = await FleetManager.coordinate({
  strategy: 'exploratory-testing',
  agents: [
    'qe-flaky-test-hunter',   // Pattern recognition
    'qe-visual-tester',       // Visual anomalies
    'qe-quality-analyzer'     // Risk assessment
  ],
  topology: 'mesh'
});

Pairing Patterns

Related Skills

• context-driven-testing - RST foundations
• risk-based-testing - Focus exploration on risk
• agentic-quality-engineering - Agent coordination

Remember

Exploratory testing = simultaneous learning, test design, and test execution.

Not random clicking. Structured, skilled investigation guided by heuristics and oracles. Document discoveries in real-time. Pair testing amplifies findings.

With Agents: Agents generate variations, recognize patterns, and maintain session notes while humans apply judgment and intuition. Combine agent thoroughness with human insight.

Gotchas

• Agent treats exploratory testing as scripted test execution — remind it: exploration means learning + adapting in real-time
• Session notes from agents lack the "why I tried this" reasoning — explicitly ask for decision rationale
• 90-minute sessions cause context overflow — cap at 45 minutes with explicit debrief
• Agent defaults to happy-path exploration — explicitly assign "Bad Neighborhood" or "Saboteur" tours for negative testing
• SFDIPOT heuristics are misapplied when agent doesn't understand the domain — provide domain context upfront