Every tool we ship targets a concrete inefficiency in how AI agents consume compute. No dashboards, no configuration, no fluff.
Applies ALL compatible skills per command. Context fix + warning ranking + RTK compression compound on every call.
Detects Jujutsu repos (.jj/) and auto-translates git commands. git add . saves ~40 tokens per call.
30+ rewrite rules reduce verbose commands to token-minimal equivalents. Community-driven.
Hook error? PRECC exits 0 instantly. Claude Code is never blocked or delayed. Zero risk to your workflow.
One curl command installs and registers the hook. Works immediately with existing Claude Code sessions.
Automated digest every Monday with last week's usage, savings, and top wasteful patterns found.
Based on your actual session data, mail-agent surfaces the 3 highest-ROI tuning actions for your workflow.
Per-project and per-session cost attribution. Know exactly where your API budget is going.
Optimised rewrites for pytest, pip, and venv commands. Saves ~18% on Python projects.
cargo-specific compression and context injection. Prevents the most common Rust compile retries.
Docker and compose command rewrites. Fixes layer cache misses and volume mount errors automatically.
AWS CLI command optimisation, region injection, and profile switching. Cuts cloud-heavy session costs.
npm/yarn/pnpm rewrites, workspace-aware context, and lockfile conflict prevention.
Write a skill, submit a PR, and earn when the community installs it.