A run_checks tool that runs named, developer-defined commands (tests, typecheck, lint, build)
and returns structured results (pass/fail, exit code, parsed file/line failures) instead of raw
terminal output, so a model can drive an edit, check, fix loop. The model picks check names,
never command lines. Commands run via execFile, never through a shell.
A
run_checkstool that runs named, developer-defined commands (tests, typecheck, lint, build) and returns structured results (pass/fail, exit code, parsed file/line failures) instead of raw terminal output, so a model can drive an edit, check, fix loop. The model picks check names, never command lines. Commands run viaexecFile, never through a shell.