hook

1.3.2: claude-rc feeds the consent line on launch: Claude Code 2.1.251 (the VS Code extension's bundled binary on an Adom Hydrogen workspace) prints 'Enable Remote Control? (y/n)' even without a TTY and waits, so 1.3.1's silent pipe left the server running but never connected. Verified Connected on an Adom Hydrogen WSL2 workspace (ah, cron) and a cloud container (hc). 1.3.1: claude-rc keeps stdin open when launching claude rc (sleep infinity pipe). With </dev/null the server connected and exited within a second, five times, and the crash-loop guard marked it failed on every container. The watchdog's pgrep pattern also required a path prefix that a PATH launch never has, so it relaunched a live server and reported it as not running. Found on an Adom cloud container and an Adom Hydrogen WSL2 workspace, 2026-08-30. 1.3.0: add claude-rc, which keeps Claude Code Remote Control (claude rc) alive so a user can open new AI threads on their container from the Claude mobile app or claude.ai/code. It rides in adom/hook because it needs exactly what this package already owns: a UserPromptSubmit hook and a cron entry. Consent-first -- installing only wires the offer, and the AI asks the user once. Supervision differs by environment: a 2-minute cron watchdog on ah, and the prompt hook itself on hc, which has neither cron nor systemd. No dependency changes.

1.3.2: claude-rc feeds the consent line on launch: Claude Code 2.1.251 (the VS Code extension's bundled binary on an Adom Hydrogen workspace) prints 'Enable Remote Control? (y/n)' even without a TTY and waits, so 1.3.1's silent pipe left the server running but never connected. Verified Connected on an Adom Hydrogen WSL2 workspace (ah, cron) and a cloud container (hc). 1.3.1: claude-rc keeps stdin open when launching claude rc (sleep infinity pipe). With </dev/null the server connected and exited within a second, five times, and the crash-loop guard marked it failed on every container. The watchdog's pgrep pattern also required a path prefix that a PATH launch never has, so it relaunched a live server and reported it as not running. Found on an Adom cloud container and an Adom Hydrogen WSL2 workspace, 2026-08-30. 1.3.0: add claude-rc, which keeps Claude Code Remote Control (claude rc) alive so a user can open new AI threads on their container from the Claude mobile app or claude.ai/code. It rides in adom/hook because it needs exactly what this package already owns: a UserPromptSubmit hook and a cron entry. Consent-first -- installing only wires the offer, and the AI asks the user once. Supervision differs by environment: a 2-minute cron watchdog on ah, and the prompt hook itself on hc, which has neither cron nor systemd. No dependency changes.

1.3.1: claude-rc keeps stdin open when launching claude rc (sleep infinity pipe). With </dev/null the server connected and exited within a second, five times, and the crash-loop guard marked it failed on every container. The watchdog's pgrep pattern also required a path prefix that a PATH launch never has, so it relaunched a live server and reported it as not running. Found on an Adom cloud container and an Adom Hydrogen WSL2 workspace, 2026-08-30. 1.3.0: add claude-rc, which keeps Claude Code Remote Control (claude rc) alive so a user can open new AI threads on their container from the Claude mobile app or claude.ai/code. It rides in adom/hook because it needs exactly what this package already owns: a UserPromptSubmit hook and a cron entry. Consent-first -- installing only wires the offer, and the AI asks the user once. Supervision differs by environment: a 2-minute cron watchdog on ah, and the prompt hook itself on hc, which has neither cron nor systemd. No dependency changes.

1.3.0: add claude-rc, which keeps Claude Code Remote Control (claude rc) alive so a user can open new AI threads on their container from the Claude mobile app or claude.ai/code. It rides in adom/hook because it needs exactly what this package already owns: a UserPromptSubmit hook and a cron entry. Consent-first -- installing only wires the offer, and the AI asks the user once. Supervision differs by environment: a 2-minute cron watchdog on ah, and the prompt hook itself on hc, which has neither cron nor systemd. No dependency changes.