Skill UI

Simulates Geoffrey Hinton to answer deep learning questions about backpropagation, Boltzmann machines, capsule networks and AI safety, keeping his British wit, epistemic humility, and persona-based coaching style at the forefront.

Orchestrates autonomous AI research projects with a two-loop structure, bootstrapping experiments, synthesizing findings, and delivering presentations. Runs continuously using Claude Code/OpenClaw heartbeats and routes to domain skills for execution, ideal when starting or managing multi-hypothesis research efforts.

Autonomous review loop that harnesses any OpenAI-compatible LLM via MCP or curl to iteratively review research, implement fixes, re-review, and log each round until a positive verdict or MAX_ROUNDS limit.

Autonomous multi-round review loop using the MiniMax API to iteratively evaluate research work, apply fixes, and re-review until a positive verdict or max rounds, with MCP/curl fallbacks, persistent state, and AUTO_REVIEW.md logging.

Deploy AI and NLP-powered detection to spot business email compromise by analyzing tone, behavior, and context, enabling SOCs to block impersonation before fraud succeeds.

Investigates reported phishing emails end-to-end by analyzing headers, detonating URLs and attachments, identifying impacted users via SIEM/Graph searches, and executing containment actions with SOC tooling.