A question I’ve kept coming back to over the last few weeks: How can I define lint rules for agents that can’t be enforced through static analysis?
Commit
b28e202
it depends on your harness, you can build hooks, and scripts that assess whatever rules you want, as long as they are measurable & deterministic; and you do like a pre-PR gate to trigger those checks, and your agent would react to their results.
100% I'm already using hooks etc. and I think what I really want is more static analysis via oxlint/eslint. Just built a rule that disallows overwriting styles of form input components. This is already so goood.
commit b28e202
yes, then it's all about triggering those linters and gates at the right moment; pstack isn't needed for this stuff