Skill UI

This skill enables non-interactive execution of the Antigravity CLI for coding tasks. It involves setting the working directory, checking CLI availability, and running tasks via JSON output format. Users must handle authentication separately and respect permission settings. It parses JSON status fields to determine success or failure. Safety measures include avoiding interactive modes, not handling credentials, and managing file permissions within the workspace. Use this when delegating coding work to Antigravity.

Run Claude Code non-interactively for coding tasks via the claude -p CLI with JSON output. Covers availability checks, timeouts, read-only tool defaults, and safe handling of missing authentication.

Run Codex CLI non-interactively to delegate coding tasks, parse JSONL event output, and enforce safe sandbox and authentication handling.

Run DeepSeek Harness non-interactively to delegate coding tasks to a headless agent. Covers availability checks, timeouts, prompt passing, exit codes, authentication limits, and safe read-only versus write workflows.

Run the Gemini CLI non-interactively to delegate coding tasks such as module reviews, using JSON output parsing, finite timeouts, and read-only permissions by default.

Run GitHub Copilot CLI non-interactively to delegate coding tasks. Covers availability checks, JSONL output parsing, and safe tool permission handling.

Run Hermes Agent non-interactively in one-shot mode for local coding tasks. Covers availability checks, timeout limits, exit code meanings, authentication boundaries, and safe workspace scoping.

Run Kimi Code non-interactively in a constrained Bash directory to delegate coding tasks. Covers availability checks, single-prompt JSONL execution, parsing final results, and safe handling of authentication, permissions, and sandbox escalation.

Run MiniMax Code non-interactively via the `mcode exec` command to delegate coding tasks. Sets the project working directory and a finite timeout, checks CLI availability, enforces read-only permissions by default, and returns the final answer. Use when the user wants MiniMax Code to analyze code or modify the workspace on request.

Runs the local OpenClaw main agent non-interactively via CLI with a finite timeout, JSON output, and read-only prompts. Use to delegate bounded tasks while avoiding onboarding, TUI, login, and credential handling.

Run OpenCode non-interactively for delegated coding tasks, parse JSON output, handle errors, and avoid interactive UI or credential handling.

This skill enables running the Pi agent non-interactively for repository analysis and implementation tasks. It involves setting the Bash working directory, checking Pi availability, and executing sessionless tasks using JSON mode. Users should delegate analysis or implementation to Pi by passing prompts as quoted arguments. Security measures include limiting tool access and avoiding credential handling.