HN Showcase
Back to browse
GitHub Repository

Agent skill that talks like Rocky - from Project Hail Mary; saving some tokens while vibing

9 starsPython

Rocky-Project Hail Mary agent skill that cut output tokens ~47%

by hpbyte·Apr 13, 2026·1 point·0 comments
Visit ProjectView on HN

AI Analysis

●●SolidBig BrainNiche Gem

Persona-based prompting cuts tokens 47% without breaking code like Caveman styles do.

Strengths
  • •Benchmarks show 47% token reduction versus standard terse prompting controls on Haiku.
  • •Preserves technical precision and code integrity better than max-compression prompt styles.
  • •Works across multiple agent platforms like Claude Code and Codex easily.
Weaknesses
  • •Relies on model behavior stability; prompts may break with future model updates.
  • •Novelty wears off quickly if token savings diminish over long coding sessions.
Target Audience

Developers using AI coding agents

Post Description

I made Rocky (from Project Hail Mary) an agent skill, similar to Caveman.

It's not maximum compression like caveman-style prompting. The goal is to keep the interaction fun and recognizable while still saving a fair amount of tokens.

Similar Projects

AI/ML●●●Banger

Freu CLI – Cut web agent token usage by 90% via compiled browser skills

Compiles browser sessions into deterministic skills, slashing agent token costs by 90%.

Big BrainSolve My Problem
0xintelligence
461mo ago
AI/ML●●Solid

Token Saving Tinyscreenshot Skill

4x token savings on screenshots with readable text at 800px grey.

Solve My ProblemBig Brain
franze
211mo ago