Score Any Public Repository Reproducibly with harness-maturity-analysis
If you are evaluating how well a repository supports AI-assisted development, it is easy to mix two different jobs: measuring one repository quickly and adding a research observation to a maintained corpus. Those jobs need different levels of evidence. harness-maturity-analysis gives you both paths. Its corpus workflow pins repositories to exact commits and records scanner versions. Its ad hoc…
Evaluating a repository's support for AI-assisted development can be done using harness-maturity-analysis. This tool provides two paths - a corpus workflow for maintaining a maintained corpus, and an ad hoc command for inspecting any local path or public repository. This tutorial focuses on the ad hoc path.
To run a deterministic local score, first clone the project using git clone, then install its development dependencies with npm install. Finally, run the documented score script with the repository URL or local path using npm run score. The command will print a maturity level, point total, dimension percentages, detected tooling, and unmet checks.
To use harness-maturity-analysis, you need Git, Node.js 18 or newer, and npm. A network connection is required when passing a remote repository URL to obtain the source and pinned scanner package. The project is MIT licensed, so read the repository license before incorporating the scripts into another workflow.
When running an ad hoc score, pass either a GitHub URL or a local directory. The ad hoc runner will obtain the target, send it to the scanner, and analyze the files without installing dependencies or executing code from the repository being inspected. This boundary is important when inspecting unfamiliar targets.
The output of the score includes a maturity level, point total, dimension percentages, detected tooling, and unmet checks. The maturity level is a compact interpretation of the score, while the point total explains how much of the available model the repository satisfies. Dimension percentages show which aspects of the repository contribute most to the result.
The detected-tooling line is descriptive and does not certify effective tool usage by the team. The unmet checks are a prioritization aid and do not automatically create a backlog.
To reproduce the observation, record the repository URL, commit SHA or tag inspected, scanner version, command and options used, complete output, and the date and environment. Rerun the same command at the same commit to ensure consistency. The corpus workflow formalizes this process by storing exact repository commits and the scanner version in corpus/manifest.json.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
