Skill UI

SAELens provides a framework for training and analyzing Sparse Autoencoders (SAEs). SAEs decompose the dense, often polysemantic activations of large language models into sparse, monosemantic features. Use this when you need to discover the discrete, interpretable concepts a model has learned, study feature superposition, or analyze specific safety-relevant behaviors (like bias or deception) within deep neural networks.

A robust, production-tested pattern for coordinating multiple specialized AI agents. The orchestrator decomposes complex tasks, routes them to the most suitable specialist, prevents redundant or conflicting work (anti-duplication), and enforces mandatory quality gates (e.g., running tests, checking file diffs) to verify results before marking a task as complete. It also includes heartbeat monitoring for task progress.

A three-tier model team orchestrator that decomposes oversized tasks into parallel subtasks, dispatches cheap workers in batches, gets plan review from a stronger advisor model, verifies each result, and synthesizes the final deliverable.

Transforms a detailed creative brief into a complete, playable song track. Users provide specifications including genre, mood, BPM, lyrics structure, and theme. The skill utilizes advanced AI tools to compose the music, structure the verses and choruses, and generate the final audio file, suitable for content creators and music prototyping.