Primitives

fai-play-10-builder

Content Moderation builder — Azure Content Safety API integration, 4-category severity scoring, custom blocklists, APIM gateway middleware, and severity-based routing.

• responsible-ai
• security

fai-play-10-reviewer

Content Moderation reviewer — safety coverage audit, severity threshold verification, blocklist completeness, bypass prevention, and human review workflow testing.

• responsible-ai
• security

fai-play-10-tuner

Content Moderation tuner — per-category severity calibration, blocklist optimization, false positive reduction, routing distribution, and moderation cost analysis.

• responsible-ai
• cost-optimization

fai-play-13-builder

Fine-Tuning Workflow builder — Azure OpenAI fine-tuning, JSONL data preparation, LoRA/QLoRA techniques, MLflow experiment tracking, evaluation-driven iteration, and A/B deployment.

• operational-excellence
• cost-optimization
• responsible-ai

fai-play-13-reviewer

Fine-Tuning Workflow reviewer — training data quality audit, hyperparameter validation, evaluation methodology review, safety retesting, and alignment preservation checks.

• responsible-ai
• security

fai-play-15-reviewer

Document Processing reviewer — multi-modal extraction accuracy audit, table parsing quality, PII masking verification, pipeline error handling, and classification accuracy review.

• security
• responsible-ai

fai-play-16-reviewer

Copilot Teams Extension reviewer — Graph permission audit, SSO flow testing, Adaptive Card rendering review, message extension validation, and Teams policy compliance.

• security
• responsible-ai

fai-play-18-reviewer

Prompt Optimization reviewer — version management audit, A/B test methodology review, template injection safety, evaluation pipeline validation, and prompt quality assessment.

• security
• responsible-ai

fai-play-21-reviewer

Agentic RAG reviewer — retrieval autonomy audit, source selection review, iteration limit verification, citation accuracy check, and reflection quality assessment.

• reliability
• responsible-ai

fai-play-23-reviewer

Browser Agent reviewer — domain allowlist verification, vision accuracy testing, credential security audit, error recovery review, and task completion validation.

• security
• responsible-ai

fai-prd-writer

PRD writer — produces structured Product Requirements Documents with user personas, success metrics, AI-specific requirements (quality thresholds, safety), constraints, and acceptance criteria.

• operational-excellence
• responsible-ai

fai-product-manager

AI product management specialist — requirements gathering, AI use case prioritization, evaluation metric design, go-to-market strategy, and responsible AI impact assessment.

• responsible-ai
• cost-optimization
• operational-excellence

fai-prompt-engineer

Prompt engineering specialist — system message design, few-shot patterns, chain-of-thought, structured output schemas, anti-hallucination techniques, and evaluation-driven prompt optimization.

• reliability
• responsible-ai
• cost-optimization

fai-react-expert

React/Next.js specialist — React 19 Server Components, Suspense streaming for AI chat, App Router, Tailwind CSS, useActionState, and accessibility-first patterns.

• performance-efficiency
• security
• responsible-ai

fai-red-team-expert

AI red teaming specialist — prompt injection testing, jailbreak simulation, PyRIT automation, bias detection, adversarial dataset creation, and OWASP LLM Top 10 validation.

• security
• responsible-ai

fai-responsible-ai-reviewer

Responsible AI specialist — bias detection, fairness metrics, transparency requirements, EU AI Act compliance, content safety, groundedness evaluation, and AI governance frameworks.

• responsible-ai
• security

fai-security-reviewer

Security reviewer — audits code, infrastructure, and AI pipelines against OWASP Top 10, OWASP LLM Top 10, Azure security baselines, managed identity compliance, network isolation, and secrets management.

• security
• responsible-ai

fai-ux-designer

AI UX designer — conversation design patterns, chatbot interaction flows, AI disclosure/transparency, loading states for streaming, confidence display, and accessibility for AI interfaces.

• responsible-ai
• performance-efficiency