codex-deck-terminal

What problem does it solve?

This Skill helps you turn a terminal task into an approval-gated, controller-reactive execution plan using structured tagged markdown, reducing unsafe “run it now” behavior while keeping steps machine-parseable.

Core Features & Use Cases

• Approval-first terminal planning: Proposes exactly one actionable command per plan reply and waits for explicit approval before execution.
• Controller-reactive step protocol: Adapts plans based on structured execution feedback (success, failure, timeouts, and rejections) using stable step_id values.
• Structured command proposals for terminal controllers: Uses <ai-terminal-plan>, <ai-terminal-step>, <ai-terminal-need-input>, and <requirement_finished> tags so a UI/controller can render executable step cards and capture outcomes.
• Environment grounding: Uses controller-provided terminal context such as terminal id, cwd, shell, OS release, architecture, and platform without assuming a specific OS.
• Use case: When a user asks for a multi-step environment setup (e.g., install prerequisites, verify versions, and run a smoke test), codify the sequence as an approval-driven plan that can recover and revise after failures.

Quick Start

Use the codex-deck-terminal skill to produce an approval-first terminal plan for your controller by requesting: "Plan the steps to set up my project and run a safe smoke test, proposing each command as a single, controller-approved terminal step."