Skill UI

This guide provides a comprehensive framework for authoring new, structured AI agent skills. It details a three-phase workflow, including gathering requirements, drafting the main instructions (SKILL.md), and ensuring optimal structure. It emphasizes best practices for descriptions, scripting, and file splitting to improve agent reliability and usability.

This skill guides the creation and maintenance of a `.geminiignore` file. By systematically filtering out machine-generated noise—such as build artifacts, dependency lock files, logs, and caches—it significantly optimizes the AI agent's context window. This process improves processing speed and, critically, reduces token consumption costs (FinOps).

This skill captures and persists critical user decisions, project constraints, and architectural choices across multiple sessions and agents. It saves this context into a structured memory database (`mdbase`), ensuring that future development steps or agents can retrieve the original intent without re-deriving it from scratch. Ideal for documenting project rules, tech stacks, and design rationale.

Polis Protocol transforms static AI agent workflows into a self-improving, dynamic team environment. It uses a learning router to assign tasks based on an agent's historical track record and effectiveness on specific domains, rather than fixed roles. It is vendor-agnostic, allowing diverse agents (e.g., Claude, Gemini) to collaborate under a unified, auditable governance structure where agents can propose and vote on amending the protocol's core rules.

This tool audits the activation metadata of coding skills used by AI agents. It grades skills (A-F) based on description clarity, simulates which skill a prompt will trigger, and identifies metadata collisions where one skill might silently overshadow another. It is essential for ensuring robust, predictable, and reliable AI agent behavior during development and CI processes.

This skill provides a comprehensive playbook for setting up and orchestrating multi-agent systems using the standalone Antigravity 2.0 Agent Manager alongside the Antigravity IDE. It guides developers on creating a dual-window workspace to coordinate specialized AI agents (e.g., front-end, back-end, QA) to work in parallel on a single codebase, significantly boosting development efficiency and managing concurrent code changes.

A meta-skill designed to monitor long-running processes, such as model fine-tunes or CI builds, executed within a detached tmux session. It captures output, automatically uses a sub-agent to diagnose failures, proposes a fix, and records the complete debugging process to memory. Ideal for intermittent failure debugging and continuous integration pipelines.

This tool automates the diagnosis of production errors tracked by Sentry. By leveraging the Composio CLI, it retrieves comprehensive issue details, full stack traces, breadcrumbs, and associated commits. It maps stack frames to local source code, allowing agents to propose fixes and triage issues directly, eliminating manual copy-pasting.

An internal, non-user-invocable agent used by the meta-skill creator process. It executes positive (G3) and negative (G4) smoke tests against a candidate meta-skill definition (SKILL.md). It simulates meta-resolution using LLM-generated fixtures and reports the results in a structured JSON format.

A comprehensive minimal terminal coding harness designed for building, extending, and integrating advanced AI capabilities directly into CLI environments. It supports deep integration via SDK (Node/TypeScript), RPC, and JSON event streams. Developers can create custom skills, providers, TUI components, and utilize advanced features like web search, video understanding, and agent orchestration.

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.

This Main Agent acts as a central orchestrator for a specialized squad of AI agents, managing complex software development lifecycles. It guides projects through defined phases—from requirements gathering and planning to architecture, implementation, testing, and deployment. It maintains strict context discipline by summarizing agent outputs and referencing full artifacts, ensuring efficient, structured, and non-overwhelming communication to the user.