kt-is-is-not - SKILL.md Agent Skill

skill_id: kt_is_is_not name: kt-is-is-not description: Kepner-Tregoe IS/IS NOT analysis for isolating the cause of deviations from normal operation version: 1.0.0 category: analysis triggers: - "сломалось" - "перестало работать" - "отклонение" - "раньше работало" - "дефект" - "причина неизвестна" - "investigation" dependencies: []

🔍 Kepner-Tregoe IS/IS NOT Analysis

Transitional reference only. The executable top-level runtime skill now lives at triz/kt_is_is_not/.

Purpose: Systematically isolate the most probable cause of a deviation from normal operation by comparing where the problem IS versus where it IS NOT.

Origin: Charles Kepner & Benjamin Tregoe (1958). Used by Ford (8D), VDA (automotive), IATF 16949.

When to use:

Something broke/failed/degraded and the cause is unknown
System worked before and now doesn't (deviation from norm)
Need to distinguish symptoms from root cause
Multiple possible causes and need to narrow down quickly

When NOT to use:

Innovation tasks (no "norm" to deviate from)
Optimization/cost reduction (use ARIZ or Value Analysis)
Vague user complaints without observable symptoms (use Empathy Probe)

Instructions

Step 1: Problem Statement

Write a clear deviation statement:

[OBJECT] has [DEVIATION] since [TIME/EVENT].
Expected: [normal behavior]
Actual: [observed behavior]

Rule: The statement must describe a CHANGE. If the system never worked correctly, KT is the wrong method.

Step 2: IS/IS NOT Matrix

Build the contrast matrix across 4 dimensions:

Dimension	IS (problem present)	IS NOT (problem absent)	Distinction
WHAT — What object has the defect?	[affected objects]	[similar objects WITHOUT defect]	[what's unique about affected?]
WHERE — Where on the object? Where geographically?	[locations with defect]	[locations WITHOUT defect]	[what's unique about those locations?]
WHEN — When first observed? When in cycle/process?	[times when defect occurs]	[times when defect does NOT occur]	[what changed at that time?]
EXTENT — How many? How much? Trend?	[scope/quantity/trend]	[expected scope if worse]	[why this extent and not more?]

Step 3: Distinction Analysis

For each row, identify the DISTINCTION — what is unique about the IS column compared to IS NOT.

Key question: "What is different, unique, or peculiar about the IS compared to the IS NOT?"

Step 4: Change Analysis

For each distinction, ask: "What CHANGED in, on, around, or about this distinction?"

List all changes that correlate with the timing of the deviation.

Step 5: Most Probable Cause

Test each possible cause against ALL IS/IS NOT data:

IF [proposed cause] is true:
  ✅ Does it explain WHY the problem IS in [IS column]?
  ✅ Does it explain WHY the problem IS NOT in [IS NOT column]?
  ✅ Does it explain the EXTENT?
  ✅ Does it explain the TIMING?

Rule: A cause must explain BOTH the IS AND the IS NOT. If it only explains IS but not IS NOT → reject it.

Step 6: Handoff to ARIZ

Once the most probable cause is identified, translate it to TRIZ language:

CAUSE FOUND: [specific cause]
→ OBJECT (Изделие): [what is affected]
→ TOOL (Инструмент): [what causes the effect — the cause itself]
→ Suggested TP: IF [cause is present], THEN [harmful effect], BUT [useful function it also serves]

Pass this to @problem_formulator for full ARIZ-85B formulation.

Anti-Patterns

Don't	Do Instead
Skip IS NOT column	IS NOT is where the insight lives
Accept first plausible cause	Test against ALL four dimensions
Use for innovation tasks	KT works only for deviations from an existing norm
Mix symptoms with causes	Symptoms go in WHAT IS, causes come from analysis
Rush to the fix	Complete the full matrix before proposing causes

Output Format

## 🔍 KT IS/IS NOT Report

### Problem Statement
[OBJECT] has [DEVIATION] since [TIME].

### IS/IS NOT Matrix
| Dimension | IS | IS NOT | Distinction | Change |
|-----------|----|----|-------------|--------|
| WHAT | ... | ... | ... | ... |
| WHERE | ... | ... | ... | ... |
| WHEN | ... | ... | ... | ... |
| EXTENT | ... | ... | ... | ... |

### Tested Causes
| # | Proposed Cause | Explains IS? | Explains IS NOT? | Verdict |
|---|---------------|-------------|-----------------|---------|
| 1 | ... | ✅/❌ | ✅/❌ | Accept/Reject |

### Most Probable Cause
[cause] → confidence: [High/Medium/Low]

### TRIZ Translation
- Object: [...]
- Tool: [...]
- Suggested TP: [...]
→ Handoff: @problem_formulator

Gate Check

Criterion	Required	Pass?
Deviation statement describes a CHANGE	Yes	✅/❌
All 4 dimensions filled (WHAT/WHERE/WHEN/EXTENT)	Yes	✅/❌
IS NOT column is not empty	Yes	✅/❌
At least one distinction identified	Yes	✅/❌
Proposed cause tested against IS AND IS NOT	Yes	✅/❌
TRIZ translation provided for handoff	Yes	✅/❌

Next Step / Handoff

Cause found → @problem_formulator (formulate TP around the cause)
Cause unclear after matrix → @causal_analysis (deeper RCA+)
Multiple equally probable causes → @experiment_designer (design test to discriminate)