Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
What This Skill Does
Workflow for supervising OpenAI Codex coding agent sessions running in background PTY processes. Covers launching sessions, monitoring log output, intervening at stuck prompts, and resuming killed sessions.
Lets you run Codex tasks fully asynchronously and intervene programmatically without keeping an interactive terminal session open.
When to use it
- Running a large refactor job in the background while continuing other work
- Detecting when Codex is waiting for an interactive prompt mid-task
- Sending approval input to a blocked Codex session without reopening a terminal
- Recovering a Codex session after a crash or manual kill
- Tracking multiple parallel Codex tasks by session ID
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Run a full-auto Codex session to add input validation to the checkout module
- 1Launch Codex in a background PTY session with the prompt and store the returned sessionId
- 2Poll logs every few minutes using process action:log to check for progress or blockage
- 3Detect an interactive approval prompt in the log output
- 4Send 'y' via process action:submit to unblock the session
- 5Wait for task completion, then summarize changed files and notify the user
Checkout module updated with input validation, Codex session completed without manual terminal intervention
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.