Skill UI

This meta-skill is designed to automate the diagnosis of complex programming errors. When presented with a stack trace, traceback, or failing log, it orchestrates a multi-step investigation. It parses the trace, searches the repository history, checks relevant issues, and analyzes code diffs to provide a comprehensive brief for root cause analysis and debugging.

A comprehensive workflow designed to convert unstructured conversational data, meeting notes, and ad-hoc decisions into highly structured, linkable entries within Notion. This tool ensures that all captured knowledge—whether an FAQ, a 'how-to' guide, or a formal decision log—is properly documented, categorized, and linked to maintain a cohesive, searchable knowledge base for easy reuse across teams.

This stub experiment generator is designed for the meta-paper-write demo. It takes a topic phrase and deterministically produces a 20-row CSV file containing baseline and generated results. This ensures stable and reproducible results for demonstration purposes, but should not be used for real scientific data.

An internal tool designed to validate candidate meta-skills (SKILL.md) before they are registered in the system. It performs comprehensive linting checks, including structural validation (G1: parsing, reference checks, XML escaping) and a scheduler dry-run (G2) using stub executors, ensuring the skill is deterministic and functional. Output diagnostics in JSON format.

An internal utility designed to manage the entire lifecycle of meta-skill proposals. It provides atomic functions to write new skill proposals, list candidates for review, and formally accept validated proposals into the active skill set. This tool is crucial for automated skill pipeline development and version control.

An internal, non-user-invocable tool designed to validate candidate meta-skills. It runs critical G3 (positive) and G4 (negative) smoke tests by simulating meta-resolution using LLM-generated fixtures. This ensures the structural integrity and robustness of the skill in a cross-vendor environment.

An internal utility designed for deep analysis of Python tracebacks. It performs Python-specific root-cause checks, provides guidance for minimal reproducible examples using pytest, and identifies precise targets for implementing defensive coding practices (like guard clauses or schema validation). Ideal for advanced debugging scenarios.

Manages the entire lifecycle of design systems within the Stitch platform. This powerful skill allows users to retrieve project assets, synthesize design tokens (DESIGN.md) from existing screens or descriptions, create/update the system in Stitch, and finally apply these tokens consistently across multiple screens to ensure a unified and scalable brand language.

A comprehensive guide covering the canonical rules for Power BI report design. It enforces consistency across design identity, visual hierarchy, color discipline, chart selection, and accessibility. Focuses on structuring reports to answer specific questions and minimizing cognitive load, ensuring data effectiveness over mere aesthetics.

Plan experiments before data collection: choose designs, randomize, block, and generate reproducible DOE layouts to avoid confounding and pseudoreplication.

This skill allows users to mine their extensive coding history across various AI agents (Claude, Codex, Gemini, etc.) into a persistent, structured corpus. It surfaces reproducible workflows as evidence and interactively helps the user forge these past practices into reusable, installed skills across their agent ecosystem. It is designed to synthesize judgment from raw transcripts into actionable, portable skills.

Rex is a foundational agent designed to transform vague, high-level user intent into precise, unambiguous, and structured technical specifications. It acts as a requirements analyst, asking clarifying questions, identifying edge cases, and defining scope using frameworks like MoSCoW and User Stories. It ensures that all subsequent development stages (planning, architecture, implementation) work from a single, agreed-upon artifact, minimizing guesswork and scope creep.