“We have an idea.
A lot is still open.”
Map the unknowns, discuss the decisions, and test what conversation cannot settle.
A small toolkit for thoughtful engineering
Better questions.
Better software.
Give your coding agent a better way to work. Ten focused skills to turn uncertain ideas into clear decisions and verified code.
Open source. MIT licensed. Start where you are.
YOUR NEXT GOOD MOVE
Find clarityAsk the useful question.
↗hc-refineMake progressBuild what matters.
↗hc-buildCheck the outcomeShow the evidence.
↗hc-reviewA toolkit, not a mandatory sequence.
YOUR AGENT. A BETTER WORKFLOW.
Claude Code · Codex · Cursor
npx skills@latest add NeuralHerd/holycow -g -a claude-code codex cursor --skill '*'Run in your terminal. Keep the agents you use. Installation options →
THE TOOLKIT / 01—10
A skill for the work
in front of you.
Use one skill or connect a few.
You bring the judgment. HolyCow brings structure.
01
holycow-guide
Find your next step.
Tell it where you are. Get a useful next step without memorizing the toolkit.
Get oriented02
holycow-init
Give your agent the lay of the land.
Discover repository conventions, checks, and tools. Keep the useful context close.
Get oriented03
hc-compass
Keep the big questions in view.
Turn uncertainty into a living map of questions, dependencies, decisions, and evidence.
Find clarity04
hc-refine
Ask better questions first.
Challenge the idea, compare realistic options, and make the trade-offs explicit.
Find clarity05
hc-lab
Make the unknown inspectable.
Test a question with a focused prototype, experiment, or diagram before committing.
Find clarity06
hc-plan
Make the work deliverable.
Turn accepted decisions into a concise brief and useful delivery increments.
Ship with confidence07
hc-build
Build it. Check it. Show the evidence.
Implement scoped work, investigate failures, and verify the behavior that matters.
Ship with confidence08
hc-review
Catch the problems that matter.
Review actual changes against intent, correctness, and project standards.
Ship with confidence09
hc-pr
Help reviewers see the change.
Explain what changed, why it matters, and which checks support the result.
Ship with confidence10
hc-retro
Make the next session better.
Look at what actually happened and turn friction into a few concrete improvements.
Keep improvingSTART WHERE YOU ARE
Every task has
a different starting point.
“We have an idea.
“We know what broke.
Let’s fix it.”
Reproduce the failure, make a focused change, and verify that the fix solves it.
“The code is ready.
Give it a careful look.”
Review the actual work, then give reviewers the context and evidence they need.
Built in the open. Still learning. These are ten foundational skills in an early release. Live workflow evaluation across all three agents is still pending. Our evaluation scenarios describe expected behavior, not passing results.
Inspired by the work of Matt Pocock, Lauren Tan, and Addy Osmani. Read the credits →