You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
teamai mcp inject --dry-run can still write config: it loads the scope without { dryRun }, so a pending legacy role migration (or a partition rename, or a self-mode bootstrap) is saved even though nothing should change.
#866 threaded { dryRun } through the loaders pull, push, status and list use; mcp inject was left out. mcp list (:27) is read-only and could load with dryRun: true unconditionally, as #866 does for status and list.
Found while merging main (#866) into #880, which threads { dryRun } through the env commands and env exec the same way (with tests). Evidence here is from reading the code on main; not reproduced with the CLI.
Description
teamai mcp inject --dry-runcan still write config: it loads the scope without{ dryRun }, so a pending legacy role migration (or a partition rename, or a self-mode bootstrap) is saved even though nothing should change.#866 threaded
{ dryRun }through the loaderspull,push,statusandlistuse;mcp injectwas left out.mcp list(:27) is read-only and could load withdryRun: trueunconditionally, as #866 does forstatusandlist.Expected:
mcp inject --dry-runwrites nothing, config included.Proposed fix: the diff above, plus a
['mcp inject --dry-run', …]row in theLOAD_ONLY_COMMANDStable the #866 tests use, so the regression is pinned.Reproduction
config.yamlis due for the legacy role migration (the fixture fix(dry-run): thread { dryRun } through the loaders pull, push, status and list use #866's tests use).teamai mcp inject --dry-run.config.yamlhas been rewritten.Environment
mainat b3b3a0bLogs
Where it was found
Found while merging
main(#866) into #880, which threads{ dryRun }through theenvcommands andenv execthe same way (with tests). Evidence here is from reading the code onmain; not reproduced with the CLI.