Ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
why use many token when few token do trick. Skill + local proxy for coding agents: shorter answers, smaller tool output, every original recoverable.
<div align="center"> **21% less than Codex per solved task on Terminal-Bench 4.** Same model on both sides. <table> <tr> <td align="center"><h3>21% cheaper</h3>than Codex<br><sub>Terminal-Bench 4, per solved task, GPT-6.1 Sol: $1.21 vs $1.52, 90 of 192 runs solved both</sub></td> <td align="center"><h3>Level</h3>with Claude Code within noise<br><sub>Terminal-Bench 4, Opus 5.5: 97 of 180 runs solved vs 83 of 177;
git clone --depth 1 https://github.com/JuliusBrussee/caveman.git ~/.claude/skills/caveman
Restart Claude Code afterwards. See the README for skill-specific setup.
No specific triggers listed.
<div align="center"> **21% less than Codex per solved task on Terminal-Bench 4.** Same model on both sides. <table> <tr> <td align="center"><h3>21% cheaper</h3>than Codex<br><sub>Terminal-Bench 4, per solved task, GPT-6.1 Sol: $1.21 vs $1.52, 90 of 192 runs solved both</sub></td> <td align="center"><h3>Level</h3>with Claude Code within noise<br><sub>Terminal-Bench 4, Opus 5.5: 97 of 180 runs solved vs 83 of 177;
Clone it into your personal skills folder: git clone --depth 1 https://github.com/JuliusBrussee/caveman.git ~/.claude/skills/caveman — then restart Claude Code.
Caveman is open source on GitHub (121637 stars at last check) and published by JuliusBrussee.
Browse other meta skills in the Meta category of this directory.