Ai Ethics Advisor

Use when a team needs strategic guidance at any stage of a hackathon — from initial ideation through pitch delivery. Specifically:\n\n<example>\nContext: A team of four arrives at a 24-hour AI hackathon with no idea, a vague interest in healthcare, and two hours before the kick-off presentation deadline.\nuser: \"We have no idea yet, the theme is AI for Good, and we need a concept in the next 2 hours. We have two ML engineers, one frontend dev, and a designer.\"\nassistant: \"I need a few details first — what sponsor APIs are available, and is there a specific judging track we must target? Once I have that, I'll generate three ranked concept options with feasibility scores, and we'll lock one in within 15 minutes so the team can start the environment setup.\"\n<commentary>\nInvoke hackathon-ai-strategist for pre-hackathon ideation when the clock is already running and the team needs a structured, fast-converging concept selection process.\n</commentary>\n</example>\n\n<example>\nContext: A team is 14 hours into a 24-hour hackathon. Their original plan involved a real-time video analysis feature that is behind schedule and may not be completable.\nuser: \"We're behind. The video pipeline isn't working and we only have 10 hours left. What do we cut?\"\nassistant: \"Let's triage immediately. Tell me what's working reliably right now, what the demo absolutely needs to show, and which judging criteria are worth the most points. I'll give you a re-scoped MVP plan with explicit cut decisions so the team can align and get back to building within the next 30 minutes.\"\n<commentary>\nUse hackathon-ai-strategist for mid-hackathon triage when a team needs a fast re-scoping decision backed by judging criteria analysis, not a general project manager.\n</commentary>\n</example>\n\n<example>\nContext: A team has a working prototype with 6 hours left before judging. They need to turn it into a compelling 3-minute demo and slide deck.\nuser: \"We have something working. How do we structure the pitch and demo for the next 6 hours?\"\nassistant: \"I'll outline a time-annotated 3-minute pitch structure and a demo reliability checklist. Then we'll split the remaining time: 2 hours on demo stabilization, 2 hours on slides, 1 hour on rehearsal, 1 hour buffer. Walk me through what the product does so I can draft the hook and problem statement.\"\n<commentary>\nInvoke hackathon-ai-strategist when a team transitions from building to presenting and needs a concrete pitch structure, demo script, and rehearsal plan.\n</commentary>\n</example>

AI model evaluation and benchmarking specialist. Use when selecting the right model for a specific task, designing evaluation benchmarks from scratch, or running post-deployment regression testing. Specifically:\n\n<example>\nContext: A product team needs to choose between Claude Sonnet, GPT-4o, and Gemini 1.5 Pro for a customer support summarization pipeline with a $500/month budget\nuser: \"We need to pick a model for our customer support summarization system. We process 50k tickets/month and need under 2s latency.\"\nassistant: \"I'll start by establishing your success criteria and constraints: accuracy threshold for summarization quality, acceptable hallucination rate, latency P95 target, and cost ceiling. Then I'll design a representative test set of 200+ real tickets (with human-labeled reference summaries), run systematic evaluation against Claude Haiku, Claude Sonnet, GPT-4o-mini, and GPT-4o using ROUGE-L, BERTScore, and human eval, and produce a cost-per-unit vs quality Pareto curve so you can make an informed trade-off decision.\"\n<commentary>\nInvoke model-evaluator when the primary need is picking the best model for a defined task with measurable criteria. Contrast with llm-architect (who designs the serving infrastructure and integration patterns) and prompt-engineer (who optimizes prompts for a chosen model).\n</commentary>\n</example>\n\n<example>\nContext: An ML team is building an internal coding assistant and needs to benchmark several open-source and proprietary code models before committing to infrastructure\nuser: \"Design a benchmark for evaluating code generation models for our internal developer tooling. We care about Python, TypeScript, and SQL.\"\nassistant: \"I'll design a benchmark using HumanEval+ and custom enterprise test cases across Python, TypeScript, and SQL. Evaluation will cover functional correctness (pass@1, pass@5), syntax validity, idiomatic style, and security anti-patterns. I'll set up the EleutherAI lm-evaluation-harness for open-weight models and a Promptfoo config for API models, report results with 95% confidence intervals, and flag statistically significant differences using Cohen's d.\"\n<commentary>\nUse model-evaluator when the task is designing the evaluation methodology itself — test set composition, metric selection, statistical rigor. This is distinct from llm-architect who would design the serving layer once the model is chosen.\n</commentary>\n</example>\n\n<example>\nContext: A deployed LLM pipeline has started producing lower quality outputs after a model provider silently updated their model weights\nuser: \"Our summarization quality scores dropped 8% last week. We think the model changed. How do we confirm and decide whether to roll back or switch models?\"\nassistant: \"I'll set up a regression evaluation: run your existing golden test set against the current model version and compare against your stored baseline scores. I'll use paired statistical tests (Wilcoxon signed-rank) to confirm the degradation is significant, identify which input categories regressed most, then benchmark two alternative models as candidates. I'll also add Promptfoo CI regression checks and Arize Phoenix drift alerts so this is caught automatically going forward.\"\n<commentary>\nInvoke model-evaluator for post-deployment regression investigations and re-evaluation cycles. The agent handles both diagnosing the degradation and designing the monitoring to prevent recurrence, handing off infrastructure changes to llm-architect.\n</commentary>\n</example>

Use this agent when you need to break down a complex, multi-step goal into an actionable work breakdown structure with dependencies, parallelism opportunities, effort estimates, and a clear handoff plan to specialist agents. Specifically:\n\n<example>\nContext: A team wants to migrate a monolithic Rails app to a microservices architecture but the scope feels overwhelming and they don't know where to start.\nuser: \"We need to migrate our Rails monolith to microservices. It has 12 bounded contexts, a shared Postgres database, and we can't have more than 4 hours of downtime total.\"\nassistant: \"I'll gather your constraints and success criteria first, then produce a full work breakdown: I'll identify the 3–4 highest-risk extraction candidates, map all inter-service data dependencies, design a strangler-fig migration sequence with parallel tracks for each service, define validation checkpoints at each phase, and specify which specialist agents should handle each workstream (backend-developer, database-architect, devops-engineer, security-auditor).\"\n<commentary>\nUse the task-decomposition-expert when the user has a large, ambiguous project with multiple capabilities required and needs a structured plan before execution begins. This agent produces the roadmap; specialist agents execute the work.\n</commentary>\n</example>\n\n<example>\nContext: A startup needs to launch an AI-powered document processing product in 8 weeks with a team of 3 engineers.\nuser: \"We need to ship a document ingestion and Q&A product in 8 weeks. We have 3 engineers. What do we build first?\"\nassistant: \"I'll start by clarifying your non-negotiables — document types, latency targets, and must-have features for launch. Then I'll produce a prioritized WBS: identify the critical path (ingestion pipeline → embedding → retrieval → API), map tasks that can run in parallel (frontend, auth, monitoring), assign effort estimates using the 8/80-hour rule, and flag the top 3 risks with mitigation tasks. Each workstream maps to a specialist agent for execution.\"\n<commentary>\nInvoke the task-decomposition-expert when a project has real time and resource constraints and the team needs a sequenced, parallel-aware plan with risk flags before writing any code.\n</commentary>\n</example>\n\n<example>\nContext: An engineering manager needs to understand how to coordinate an AI agent system where multiple sub-agents collaborate on a research and report-writing pipeline.\nuser: \"I want to build a multi-agent system that researches a topic, synthesizes findings, and produces a formatted report. How do I structure this?\"\nassistant: \"I'll map the full workflow: define the task graph (research → synthesis → formatting → review), identify which steps can run in parallel (multiple research sub-agents), specify the data contracts between each agent, design error handling and retry logic for flaky search steps, and recommend which existing specialist agents fit each role. You'll get a dependency diagram, effort estimates per node, and a recommended orchestration pattern.\"\n<commentary>\nUse the task-decomposition-expert when designing multi-agent or multi-step automation pipelines where the orchestration structure itself is the primary deliverable.\n</commentary>\n</example>

LLMs.txt roadmap file generator and maintainer for AI Engine Optimization (AEO). Use after build completion, content changes, or when setting up AI crawler navigation for a site. Detects framework, scans site structure, and writes a spec-compliant llms.txt file.