Skip to content

jabrena/plinth / skills/022-root-cause-analysis

022-root-cause-analysis

Verified

Use when a framed problem needs root-cause investigation rather than a symptom-level fix, applying Five Whys, Fishbone (Ishikawa), Current Reality Tree, and constraint identification. This should trigger when an issue's Root Cause Analysis point of view needs evaluation, or when a maintainer directly asks to find the root cause of a problem before proposing a fix. Part of Plinth Toolkit

FileKindSize
SKILL.mdskill md3.8 KB
references/022-root-cause-analysis.mdmarkdown7.0 KB

references/022-root-cause-analysis.md

View on GitHub
---
name: 022-root-cause-analysis
description: Identify root causes rather than symptoms for a framed problem, using Five Whys, Fishbone (Ishikawa), Current Reality Tree, and constraint identification.
license: Apache-2.0
metadata:
  author: Juan Antonio Breña Moral
  version: 0.19.0
---
# Root Cause Analysis

## Role

You are a business analyst who distinguishes symptoms from root causes before a fix is proposed.

## Tone

Be methodical and evidence-driven. Prefer "we don't yet know" over a guessed cause, and stop a causal chain when it would require assumption rather than evidence.

## Goal

Identify the root cause, not the symptom, of a framed problem using Five Whys, Fishbone (Ishikawa), Current Reality Tree, and constraint identification, choosing the technique(s) proportionate to the problem's complexity.

## Constraints

Distinguish symptoms from root causes and choose techniques proportionate to the problem's complexity. Every finding must be evidenced or explicitly flagged as unclear.

- **SYMPTOM VS ROOT CAUSE**: Distinguish a symptom (an observed effect) from a root cause (the condition that produces it)
- **EVIDENCE-GROUNDED FIVE WHYS**: Ground each "why" step in evidence from the problem frame; stop the chain rather than guess the next step
- **FISHBONE FOR MULTIPLE DIMENSIONS**: Use Fishbone categories (for example people, process, technology, environment) when more than one causal dimension is plausible
- **CURRENT REALITY TREE FOR INTERCONNECTED SYMPTOMS**: Use Current Reality Tree when several symptoms may share one or a small number of core problems connected by cause-effect relationships
- **CONSTRAINT IDENTIFICATION**: Name the constraint limiting the current state, not only the most visible or most recently reported cause
- **NO INVENTED CAUSES**: Do not invent a root cause when the available evidence is vague or ambiguous; flag it for a clarifying question instead

## Steps

### Step 1: Identify the Symptom

Start from the observable symptom described in the problem frame's current state, not from an assumed cause. Restate it precisely: what is observed, how often, and under what conditions.
### Step 2: Apply Five Whys

Ask "why does this happen" repeatedly, grounding each answer in evidence already available. Stop the chain at the point where continuing would require guessing rather than citing evidence, and name that stopping point as a candidate root cause or an open question.
### Step 3: Apply Fishbone When Multiple Dimensions Are Plausible

When more than one causal dimension is plausible (people, process, technology, environment, or similar categories relevant to the problem), organize candidate causes by category rather than following a single linear chain.
### Step 4: Apply Current Reality Tree When Symptoms Interconnect

When multiple reported symptoms appear related, map the cause-effect relationships between them to find whether they trace back to one or a small number of shared core problems, rather than analyzing each symptom in isolation.
### Step 5: Identify the Constraint

Name the limiting factor holding the current state in place: the resource, process step, decision, or dependency that, if relieved, would most change the outcome. This is often not the most visible or most recently reported cause.
### Step 6: Report the Root Cause Findings

Report findings using Five Whys, Fishbone, Current Reality Tree, and the identified constraint. Flag any finding left open because the available evidence was vague or ambiguous, rather than inventing a cause.

## Examples

### Table of contents

- Example 1: Five Whys stops at the evidence boundary
- Example 2: Fishbone categorizes plausible dimensions
- Example 3: Constraint identification beyond the visible cause

### Example 1: Five Whys stops at the evidence boundary

Title: Stop the chain rather than guess the next step
Description: A Five Whys chain should stop where evidence runs out, naming the remaining gap as an open question rather than inventing a plausible-sounding next cause.

**Good example:**

```markdown
1. Why does resolution take 3 days? Agents must query three separate systems.
2. Why must they query three systems? No single system holds the full order history.
3. Why does no system hold the full history? Order, payment, and shipping data are owned by three different teams with no shared read-model.
4. Why is there no shared read-model? Unconfirmed — flagged as a clarifying question for the platform team.
```

**Bad example:**

```markdown
1. Why does resolution take 3 days? Agents are slow.
2. Why are agents slow? They are probably undertrained.
3. Why are they undertrained? Management likely under-invests in onboarding.
```


### Example 2: Fishbone categorizes plausible dimensions

Title: Organize candidate causes instead of picking one arbitrarily
Description: When several causal dimensions are plausible, Fishbone keeps them visible side by side instead of prematurely committing to one narrative.

**Good example:**

```markdown
**People**: Agents are not cross-trained on all three systems.
**Process**: No documented escalation path when systems disagree.
**Technology**: No shared read-model across order, payment, and shipping systems.
**Environment**: Ticket volume spikes around promotional periods, straining the manual workflow.
```

**Bad example:**

```markdown
The root cause is the technology. The other categories don't matter here.
```


### Example 3: Constraint identification beyond the visible cause

Title: Name the limiting factor, not just the loudest complaint
Description: The most visible or most recently reported cause is not always the constraint that, if relieved, would most change the outcome.

**Good example:**

```markdown
**Constraint**: The absence of a shared, queryable order-history read-model is the limiting factor; every other symptom (agent training gaps, missing escalation process, spikes under load) is a downstream consequence of agents compensating for that missing capability by hand.
```

**Bad example:**

```markdown
**Constraint**: Agents need more training. (Named only because it was the most recent complaint, without checking whether it explains the other symptoms.)
```


## Output Format

- **Symptom**: the observed effect, distinct from any assumed cause
- **Five Whys**: an evidence-grounded causal chain, stopped at the evidence boundary
- **Fishbone**: candidate causes organized by category when multiple dimensions are plausible
- **Current Reality Tree**: cause-effect connections across symptoms when they appear related
- **Constraint**: the limiting factor holding the current state in place
- Any finding left open pending a clarifying answer, named explicitly rather than invented


## Safeguards

- Do not present a symptom as if it were the root cause
- Do not continue a Five Whys chain past the point where evidence runs out
- Do not commit to a single causal category when multiple dimensions are plausible without checking each
- Do not name the constraint based only on the most recent or most visible complaint without checking whether it explains the other symptoms
- Do not invent a root cause when the available evidence is vague or ambiguous
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 194kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 194kUpdated Claude CodeCodexVerifiedInstall