lbzepoqo

Skills for how Claude works on a task: hard multi-step tasks, context docs, surviving a usage limit, and git worktrees

4 skills

Compacts the conversation at task boundaries instead of at a fixed token count: the model marks the session busy or at a boundary, and the plugin compacts at the end of a turn when the mark and the context size allow it

Output style: answer first, keep every figure, do the work you can do, and say what was checked

Points Claude to the saved auto-memory notes that match each prompt you type, so a note is read before the work it applies to instead of sitting unread behind its index line

Tells you when installed plugins or copied skills are behind upstream, without updating them: a status line, one toast per news, and a /plugin-updates list of the missing commits

Reminds Claude to run code-review after a code commit that no review saw: from 30 changed lines, or any change to guard code; adds /simplify from 300 lines

Keeps secrets out of the conversation: hides keys before Claude or the transcript sees them, and blocks credential files and commands that print secrets. Fork of arasovic/claude-code-mods

Gives Claude a search over your past Claude Code sessions: it greps the raw transcripts on demand, with no index, no model calls and nothing added at session start

Shows which installed skill fits each prompt you type, as a toast and a status line, and lists every skill in a /skill-hints pane

Shows a toast when a prompt starts a different topic than the earlier requests in the session, so you can /clear first