Push Claude Code completions to your phone via ntfy. The Stop hook sends what Claude actually said at turn-end (PR link / decision / error) instead of a static message; Notification events push permission/attention prompts. Auto-generates a private topic on first run with subscribe instructions. Cross-platform Node, no dependencies.
1 command · hooks
Ships a UserPromptSubmit hook that prepends 'Remember the code commenting rule.' to every prompt, plus a skill containing the rule itself (when to comment, when not to). Keeps Claude's commenting discipline consistent across long sessions and after context compactions, without per-seat settings.json edits.
1 skill · hooks
Stops every Claude Code agent from spending usage credits until you say so. When your claude.ai plan crosses a usage threshold (5-hour 95%, weekly 98% by default), all sessions, subagents and workflow agents pause on one macOS dialog: Allow credits until the window resets, or Stop. One-time /credit-gate:setup connects it to your status line. macOS only; needs jq.
1 command · hooks