codecast

What This Skill Does

The Codecast skill acts as a high-fidelity bridge between your terminal-based AI coding agents (such as Claude Code, Codex, or custom CLI agents) and Discord. Instead of trapping your agent's thought process inside a black-box terminal window, Codecast captures and parses the stream-json output, translating it into clean, readable Discord messages in real-time. It monitors tool calls, file writes, bash commands, and final results without consuming any additional AI tokens for the relay process. By offloading the observability of your dev sessions to Discord, you enable team transparency, remote auditing of AI behavior, and persistent logging of complex coding tasks.

Installation

To integrate Codecast into your OpenClaw environment, execute the following installation command in your terminal: clawhub install openclaw/skills/skills/allanjeng/codecast

Ensure you have configured your webhook references as outlined in the setup documentation. Before running production tasks, perform a smoke test by initiating a simple echo command through the relay to verify that your Discord channel receives the structured output correctly.

Use Cases

• Remote Debugging: Allow teammates to observe AI coding sessions in a Discord thread while you are away from the machine.
• Agent Auditing: Maintain an immutable record of all file writes and bash commands performed by an agent for security and compliance review.
• Collaborative Development: Enable pair programming with AI where the entire team can see the reasoning and execution flow in real-time.
• Asynchronous Handoffs: Use the completion detection feature to automatically report task results back to the invoking Discord channel once the agent process terminates.

Example Prompts

1. "Stream my current coding session to the #dev-updates channel using Codecast."
2. "Relay my Claude Code output to Discord and enable thread mode so my team can review the changes."
3. "Start a new dev session for the current directory and make sure the output is visible in my daily-work Discord channel."

Tips & Limitations

• Zero Token Overhead: Because the relay parses output streams directly, you do not pay for extra LLM calls to summarize or translate the output.
• Background Persistence: Always launch with exec background:true to ensure your session remains active even if your primary interface connection drops.
• Completion Detection: Utilize the breadcrumb file pattern (/tmp/codecast-pending-<PID>.json) to ensure automatic reporting. If you encounter issues, verify that your agent process is correctly signaling completion.
• ANSI Support: While structured JSON is best, the skill provides basic support for raw ANSI output from arbitrary CLI tools, though readability may vary compared to native Claude Code integration.