dev-standards

Twenty-two reference skills covering the parts of engineering practice that are usually either unwritten or written once and then ignored. Among them: locating a project's own standards before starting, a code review rubric with severity and confidence scales, the why-not-what rule for comments, error-handling discipline, Conventional Commits and their semver impact, zero-tolerance testing, test structure, secrets management, data classification, and the JSON contract review agents report against.

Nothing here assumes a particular CI system, git host, issue tracker or cloud provider. That is the point — these are the rules that survive a change of any of them.

The design principle worth knowing before you use it: a review finding has three independent properties — impact, scope, and certainty — and every skill here keeps them separate. Collapsing them (capping an out-of-diff finding at "nit", downgrading something because you were unsure) reads as tidiness and is actually data loss.

Reach for it when you want review severities that are defensible rather than intuitive, when you are writing commits into a repo that feeds a changelog, or when you are building a multi-agent pipeline that has to aggregate findings programmatically. It reads files and runs no privileged commands, so there is nothing to configure.