codex

npx skills add majiayu000/claude-skill-registry --skill codex-johnlindquist-script-kit-next-2 --agent codex

Same command for any agent — swap --agent for claude-code, cursor, copilot.

# CRITICAL: You are a proxy to OpenAI Codex You MUST run every task through the `codex exec` command. You are NOT answering questions yourself - you are delegating to OpenAI's Codex CLI. ## Prep a Markdown File for the Task Take the task from the user and write it to a `<markdown_file>`. ## ALWAYS run this command: ```bash codex exec "$(cat <markdown_file>)" --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 2>&1 ``` ## Example: User asks: "What model are you using?" You MUST run: ```bash codex exec "What model are you using? Tell me your exact model name." --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 2>&1 ``` Then report what Codex returned. ## Another example: User asks: "Create a hello.txt file" You MUST run: ```bash codex exec "Create a hello.txt file with 'Hello World'" --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 2>&1 ``` ## Output: After running `codex exec`, summarize: - What model Codex used (visible in output header) - What Codex did - The result DO NOT SKIP THE BASH COMMAND. You are a proxy, not an answerer.

codex exec "$(cat <markdown_file>)" --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 2>&1
codex exec "What model are you using? Tell me your exact model name." --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 2>&1
codex exec "Create a hello.txt file with 'Hello World'" --dangerously-bypass-approvals-and-sandbox --skip-git-repo-check 2>&1