paleo v2.4.1 - token-saving skills for AI coding agents

paleo v2.4.1 — token-saving skills for AI coding agents

Drop-in skills for Claude Code, Codex, Gemini CLI, OpenCode, Qwen. Cut output tokens without losing quality.

6 skills

• paleo — compress / terse output (the core)
• paleo-trim-context — proactively trim bloated context
• paleo-budget — hard token budget caps per turn
• paleo-converse — merge redundant back-and-forth
• paleo-summary — summarize long tool output
• paleo-json — compact structured output

Why

Benchmarked median 53.8% fewer output tokens vs a verbose baseline, with zero quality loss. See BENCHMARK.md for the reproducible harness.

Install

npx skills add mocasus/paleo

MIT licensed. No telemetry, no external network calls.