Guards

Forecasts how many turns remain until auto-compaction from smoothed per-turn context growth (tool output vs conversation), and at natural breaks offers Compact now with a focus Haiku drafts.

When Claude re-reads a file that changed since it last read it, it gets only the changed hunks (current line numbers, a few context lines) instead of the whole file.

Huge Bash/PowerShell outputs without the context bill: Claude gets the head and tail, the full text goes to a file it can open.

Stops Claude from re-reading a file it already read in full and that has not changed since; counts the tokens saved.

Blocks destructive shell commands and edits to secrets before they run

/dash opens a live side pane with tool calls, files touched, tokens and turn times

Tracks tokens and cost per model, agent and session for every session on a git branch, in a right pane.

Phone-driven sessions end each turn with tappable AskUserQuestion options and a push notification, injected per-turn and scoped to the session you toggle it in.

Mods for the laa toolkit: blocks file changes and commits on the default branch, shows laa status above the prompt, a pane to triage open learnings, and a /laa:retro nudge when a pipeline ends.

A pixel theater under the prompt: the little orange guy acts out what Claude is doing, with Haiku narration in a speech bubble.

Rewrites lazy commit subjects into conventional commits from the staged diff, and strips AI attribution trailers

A red band when you are on main or in prod, and no force-push, reset --hard or unconfirmed push to a protected branch

Late at night Claude gets more careful and asks before deploys, force pushes and rm -rf

Approve or deny Claude Code's permission prompts from Telegram, with the terminal prompt as the fallback.

A fast second opinion on every edit: a small model flags removed error handling, deleted tests, secrets and more

Stops Claude from writing API keys, tokens, private keys and database passwords into files.

Live WHOOP heart rate in the status line; risky Bash commands are blocked while it is over your limit

Dot-matrix HUD for Claude Code: chat tokens by model, 5-hour limit estimates, cache timer and code stats. Local only, no model calls.

Shows which files Read opened, which lines, and what each search looked for, on the folded tool line

Opinionated framed status band for Claude Code with git, model, context, and the branch's PR review state

Catches risky shell commands (rm -rf, git reset --hard, force push, ...) before they run, forces a permission prompt, and shows what they would touch in a side pane.

A live sidebar of every file Claude touched this session, with +/- line counts.

Understand a codebase visually: a live architecture map that lights up as Claude reads and edits, a call graph explorer, the session's changes by layer with their impact, a decision timeline, and side-question explanations that never enter your conversation. /atlas

Agents side pane: every subagent as a little pixel crew member with its model, effort, context, tokens, cost and time; running, completed and planned tasks with their dependencies. /crew