skills.rest
codewizwit
codewizwitCOMMUNITY
Explore all skills by this author →
GitHub

bias-detection

Analyzes code, text, and datasets for bias across protected characteristics with severity-rated findings.

1|Updated Oct 11, 2025
Search Tags:#code review#governance#fairness#bias detection#ai ethics#responsible ai#inclusivity audit
One-click install
npx skills add https://github.com/codewizwit/human-in-the-loop --skill bias-detection-codewizwit
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: bias-detection
Source: https://github.com/codewizwit/human-in-the-loop/tree/main/lib/skills/bias-detection
Command: npx skills add https://github.com/codewizwit/human-in-the-loop --skill bias-detection-codewizwit

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? AI-generated content, code, and datasets can contain hidden bias against gender, race, age, disability, cultural background, or socioeconomic status, and manual review often misses subtle or intersectional patterns. ## Core Features & Use Cases - Multi-Dimensional Analysis: Evaluates content across gender, racial, age, disability, socioeconomic, and cultural bias dimensions with intersectional detection. - Actionable Findings: Quotes specific problematic content, explains why it is biased, assigns severity ratings, and provides concrete alternative phrasing. - Use Case: Review an authentication system's source code for age-based restrictions and gendered pronoun assumptions, then receive a structured report with refactored code alternatives and a deployment readiness recommendation. ## Quick Start Ask the assistant to check your code, AI-generated text, or dataset for bias and fairness issues across demographic groups.

Frequently Asked Questions about bias-detection

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I check code for bias and fairness issues?▼

Point the analyzer at your source files or paste the code directly, then select the bias dimensions to evaluate. It scans for patterns like age-based restrictions, gendered assumptions, and discriminatory eligibility logic, returning severity-rated findings with refactored alternatives.

What types of content can be analyzed for bias?▼

The analyzer supports four content types: source code and algorithms, AI-generated text or copy, automated decisions and recommendations, and training datasets. You choose the type during an interactive intake step so the analysis matches the content.

How are bias findings rated for severity?▼

Each finding receives a severity rating of Critical, High, Medium, or Low based on potential real-world harm to affected groups. The report also includes a per-dimension summary table and an overall deployment readiness recommendation.

Can I audit a training dataset for representation bias?▼

Yes, dataset audits produce a statistical summary of demographic representation, identify underrepresented groups, analyze label distribution across demographics, and recommend dataset balancing strategies.

What are the limitations of automated bias detection?▼

Bias detection depends heavily on context, so the tool gathers content type and concern areas before analyzing. Borderline cases are flagged with reasoning rather than definitive labels, and findings should be reviewed by humans before remediation decisions.

Related Skills

Explore More Legal & Compliance Skills→
Bakery88Bakery88

comfyui-lora-data-gathering

Gather, curate, and auto-caption LoRA training data into a ready dataset folder.

Community
Advanced
realsigridjinrealsigridjin
20

code-review

Review git diffs for security, quality, and performance issues.

Community
Advanced
DTMC-marketplaceDTMC-marketplace

ai-ethics

Analyze AI models and datasets for bias and fairness metrics.

Official
Advanced
DTMC-marketplaceDTMC-marketplace

bias-assessment

Evaluate AI systems for fairness using demographic parity and equalized odds.

Official
Advanced