Skill UI

This skill is designed to configure a Rails project workspace to function correctly within Conductor, a platform for parallel coding agents. It sets up the necessary scaffolding, including `conductor.json`, executable setup scripts, and server scripts. Additionally, it updates critical Rails configuration files (such as Sidekiq, Cable, and development environment settings) to ensure proper environment variable handling, Redis isolation, and port management for reliable parallel agent execution.

Addresses the critical challenge of managing extremely long conversation histories (millions of tokens) in advanced AI agents, such as coding assistants. This technique moves beyond simple token minimization by implementing structured compression strategies (like Anchored Iterative Summarization) to prevent the loss of critical technical details—such as file paths, decisions, and error messages—ensuring the agent maintains functional context across complex, multi-step tasks. It emphasizes optimizing for 'tokens-per-task' rather than just tokens-per-request.

This advanced system transforms session interactions into reusable, atomic knowledge units called "Instincts." By hooking into tool usage and session activity, it observes patterns (like preferred coding styles, error fixes, or repeated workflows). These instincts are then clustered and "evolved" into structured Skills, Commands, or specialized Agents, capturing deep contextual knowledge with weighted confidence scoring to enhance development productivity.

This skill enables persistent, file-based planning for AI agents, acting as an external, reliable working memory on disk. By utilizing dedicated markdown files (task_plan.md, findings.md, progress.md), it allows complex, multi-step projects, research tasks, or coding sprints to survive context loss and session resets. It is essential for maintaining state and tracking progress across long-running, intensive workloads.

This comprehensive guide outlines the best practices and structured methodology for developing specialized skills for AI coding agents. It provides detailed instructions on integrating Azure SDKs and Microsoft Foundry services, covering crucial topics like secure authentication patterns (DefaultAzureCredential), modular workflow design, and maintaining concise, effective knowledge context for optimal agent performance.

Audits GitHub Actions workflows that invoke AI coding agents to surface prompt injection and sandbox risks, tracing attacker-controlled inputs through env vars, reusable actions, and workflow references before deployment or code review.

A powerful tool to scaffold and ship production-ready Model Context Protocol (MCP) servers directly from OpenAPI specifications. It supports Python and TypeScript, ensuring schema validation, safe evolution, and robust tool definition generation, eliminating the need for manual wrapper coding. Use it to expose existing REST APIs to LLM agents or build shared tool integrations.

Performs static security analysis on GitHub Actions workflows that incorporate AI coding agents (e.g., Claude, Gemini, OpenAI). This skill helps identify critical data flow vulnerabilities, such as where attacker-controlled input can reach AI agent prompts or internal systems via CI/CD pipelines, going beyond simple expression checks.

This infrastructure provides a robust solution for running powerful AI agents in remote, sandboxed environments. It overcomes the limitations of local execution by ensuring unlimited concurrency, consistent environments, and collaborative features. Use this to build complex background coding agents, multi-client interfaces (e.g., Slack, Web), or scale agent workloads far beyond local machine constraints.

A robust framework for orchestrating complex, multi-step coding tasks using multiple AI agents. It functions by first planning the entire project into a mission dependency graph (DAG), then dispatching specialized agents to isolated Git worktrees for parallel development. The system automatically monitors progress, handles merges, and generates structured, comprehensive reports detailing changes, completed tasks, and next steps.

A lightweight CLI tool for systematically comparing various coding AI agents (such as Claude Code, Aider, and Codex) head-to-head. It measures crucial metrics—pass rate, API cost, execution time, and consistency—on reproducible tasks defined via YAML manifests. Use this to objectively select the best agent for your codebase or to validate model updates before production deployment.

This skill acts as an anti-over-engineering guardrail, instructing AI coding agents to maintain strict scope control and minimalism. It guides the AI to only modify the files and lines explicitly requested by the user, avoiding unnecessary abstractions, unrequested refactors, or scope creep. Use it when you prefer the simplest, most direct solution and want to prevent the agent from 'grinding' the code base.