how/agentic by Statue Labs
A TEST FOR THE WAY AGENTS WORK

How agentic
is your repo?

See your codebase through an agent’s eyes.
Find the friction. Make the next run better.

Public GitHub repos · No sign-inSee an example

EARLY PREVIEW Live tests currently inspect the repository. Grading is next.

A LITTLE LESS FRICTION.
LESS GUESSWORK.
MORE FORWARD MOTION.
01 / THE TESTCONTEXT SHOULD BE SCULPTED.

A README can make promises.
An agent has to follow them.

The test is designed to investigate how well an agent can understand, navigate, and work in your repository. The evidence matters more than the presence of a file.

01

Get oriented.

Find the instructions, map the code, and understand where a change belongs.

INSTRUCTIONS + NAVIGATION
02

Follow the evidence.

Check documented commands. Investigate stale paths, conflicting guidance, and dead ends.

BUILD + INVESTIGATION
03

Make the next run better.

A score, a visible journey, and specific suggestions that give you somewhere to start.

REPORT + SUGGESTIONS

We’re building toward this full test. Today’s live scan validates the repo, inventories its files, and records the commit. Explore the full report design →

A few good questions.

What does the score measure?

The planned grade measures the friction an agent encounters: whether instructions agree, documented commands work, and the code is easy to navigate. Only a completed investigation receives a grade. The current inventory scan does not issue a score.

Does the test change my repository?

No. A scan checks out the public repository and inspects it in a separate workspace. The current test runs no code from your repository and makes no changes to GitHub.

What appears on the leaderboard?

Completed, graded investigations. The latest completed scan replaces the previous entry for that repository. Inventory scans and partial reports are not ranked. Example entries are labeled and kept in a separate preview.