# Improve Evidence-Based Reasoning

Complex reviews and decisions often fail when claims outrun evidence. This skill provides verification workflows, hypothesis testing, and self-correction prompts.

## Install

```bash
npx skillstore add joseph obrien/89jobrien-meta-cognitive-reasoning
```

## Metadata

- Status: approved
- Slug: 89jobrien-meta-cognitive-reasoning
- Version: 1.0.1
- Author version: 1.0.1
- Skillstore revision: r1
- Version status: valid
- Tree hash: 5faee4871ecc5829f8380667f97b1e6d11a02876c7dfc09b19375347d8c90c6b
- Author: Joseph OBrien
- GitHub username: 89jobrien
- License: MIT
- Repository: https://github.com/89jobrien/steve/tree/main/steve/skills/meta-cognitive-reasoning
- Ref: c4037264bbd363c572662d6154a3ab28f5ca4f53
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: network, external\_commands
- Quality score: 78
- Quality tier: bronze
- Public page: https://skillstore.pages.dev/skills/89jobrien-meta-cognitive-reasoning
- Manifest: https://skillstore.pages.dev/api/skills/89jobrien-meta-cognitive-reasoning/manifest

## Capabilities

- Prompts users to show evidence before making claims.
- Guides multiple hypothesis generation before conclusions.
- Adds temporal checks for package, API, and tool versions.
- Structures self-correction when earlier analysis was wrong.
- Provides review workflows for code, architecture, and documentation.

## Use Cases

- Review Code Changes: Engineers can require evidence for every finding before changing code.
- Debug Complex Failures: Support teams can compare possible causes before choosing a fix.
- Assess Technical Decisions: Technical leads can weigh assumptions, evidence, and tradeoffs before approving plans.

## Prompt Templates

### Verify a Claim

```
Use meta-cognitive reasoning to verify this claim before accepting it: [claim]. Show evidence first, then interpret it.
```

### Compare Hypotheses

```
Analyze this issue with at least three hypotheses: [issue]. Identify the evidence that would confirm or reject each one.
```

### Review With Evidence

```
Review this change for correctness and risk. Cite specific evidence for every finding, and separate observation from assessment.
```

### Recover From an Error

```
Reassess my previous conclusion about [topic]. State the claim, check evidence, explain any error, and give the corrected action.
```

## Limitations

- It does not verify facts without access to authoritative sources.
- It cannot replace domain expertise for legal, medical, or financial decisions.
- It may slow simple tasks that need quick execution.
- It relies on the agent to follow the reasoning workflow.

## Best Practices

- Use the skill before high-impact reviews, not after conclusions are formed.
- Provide files, logs, or source links so evidence can be checked.
- Ask for separate observation, mechanism, and assessment sections.

## Anti Patterns

- Asking for a final verdict without providing evidence sources.
- Using the skill to justify a decision already made.
- Skipping verification because an issue matches a familiar pattern.

## Security Audit

- Audited at: 2026-07-05T12:59:30.078\+00:00
- Summary: All static findings were adjudicated as false positives. The command alerts are Markdown examples, the URLs are PyPI registry examples, and the reconnaissance alerts are ordinary reasoning prose. No prompt injection, exfiltration intent, or executable payload was found.

## Stats

- Views: 198
- Downloads: 16
- Favorites: 0
- Popularity score: 0