# Why Agent Reliability

AI coding agents are most useful when a repository gives them a narrow, trustworthy operating surface.

That surface includes:

- clear instructions,
- repeatable commands,
- safe handling of secrets,
- CI that proves changes,
- README instructions that users can replay,
- tool permission boundaries,
- visible maintainer expectations.

Agent Reliability Kit turns those signals into a report that maintainers can use before merging AI-assisted changes or inviting contributors.

Scans are local-first: source code and findings stay on your machine unless you choose to share a report. Shared reports, examples, and issues must not include real secrets, private logs, cookies, browser profiles, or private URLs.