# Diagnose Problems With Abductive Reasoning

Teams often face incomplete data when explaining anomalies, failures, or unexpected outcomes. This skill guides Claude, Codex, and Claude Code through structured hypothesis generation, evidence scoring, and ranked explanations.

## Install

```bash
npx skillstore add bellabe/reasoning-abductive
```

## Metadata

- - Slug: bellabe-reasoning-abductive
- - Version: 1.0.0
- - Author: BellaBe
- - GitHub username: BellaBe
- - License: MIT
- - Repository: https://github.com/BellaBe/lean-os/tree/main/.claude/skills/reasoning-abductive
- - Ref: main
- - Supported tools: Claude, Codex, Claude Code
- - Risk level: low
- - Quality score: 78
- - Quality tier: bronze
- - Public page: https://skillstore.pages.dev/skills/bellabe-reasoning-abductive
- - Manifest: https://skillstore.pages.dev/api/skills/bellabe-reasoning-abductive/manifest

## Capabilities

- Structures raw observations into clear anomaly descriptions.
- Generates multiple plausible explanations across technical, product, market, operational, and external categories.
- Scores hypotheses using explanatory power, simplicity, coherence, testability, and prior probability.
- Ranks candidate explanations with confidence levels and supporting evidence.
- Identifies contradicting evidence, uncertainty, and practical next steps.

## Use Cases

- Diagnose product metric drops: Explain changes in conversion, activation, retention, or revenue by comparing competing hypotheses against available evidence.
- Investigate operational anomalies: Turn unexpected process failures or performance changes into ranked causes and practical verification steps.
- Analyze incident symptoms: Generate likely explanations for system behavior from logs, timelines, impact scope, and known changes.

## Prompt Templates

### Explain an anomaly

```
Use abductive reasoning to explain this observation: [describe anomaly]. Generate at least five hypotheses and rank the top three.
```

### Compare possible causes

```
Analyze these observations and evidence: [data]. Score each hypothesis for explanatory power, simplicity, coherence, testability, and prior probability.
```

### Find missing evidence

```
Use abductive reasoning on this issue: [issue]. Identify the strongest explanations, contradicting evidence, missing evidence, and tests to reduce uncertainty.
```

### Build a full diagnostic brief

```
Create a complete abductive reasoning brief for [case]. Include anomaly framing, diverse hypotheses, evidence scoring, ranked conclusions, confidence bounds, and next actions.
```

## Limitations

- It does not collect evidence automatically from external systems.
- Its conclusions depend on the quality and completeness of provided observations.
- It cannot prove causality without follow-up tests or additional data.
- It may produce several plausible explanations when evidence is sparse.

## Best Practices

- Provide concrete observations, timeframes, baselines, and affected segments.
- Include both supporting and contradicting evidence for each major hypothesis.
- Treat the top explanation as provisional until follow-up tests confirm it.

## Anti Patterns

- Do not stop after the first plausible explanation.
- Do not use confidence scores when evidence is missing or vague.
- Do not treat abductive reasoning as proof of causality.

## Security Audit

- - Safe to publish: true
- - Audited at: 2026-06-28T15:01:04.23\+00:00
- - Summary: Static analysis reported shell execution, weak cryptography, and reconnaissance patterns, but contextual review found only Markdown examples, ordinary prose, and structured reasoning templates in SKILL.md. No executable commands, prompt injection attempts, network access, filesystem access, credential handling, or malicious intent were found.

## Stats

- - Views: 225
- - Downloads: 7
- - Favorites: 1
- - Popularity score: 0