audit: agreement true says nothing about whether in-flight work is healthy #41
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Environment: concord v0.9.0, registry version 3, Linux. Observed while sweeping 21 tasks across 3 domains.
Observed
Every task in the sweep reported
agreement: true to the protocol. In the same sweep:MAIN.mdclaimed the implementation was complete;MAIN.mdsaid its work was committed but not landed, while the commit was alreadyorigin/main;origin/mainwith an open, conflicting draft PR untouched for seven days.None of that is visible from
audit. On a member with 22 dirty files today:Why this is worth changing
Four- (now five-) surface agreement is the right gate for mutation, and I am not asking to widen it. But
auditis also the documented task-entry command inprotocol.md"Entering and starting" step 2. An operator entering a task readstrueand moves on.v0.9.0 already has the extension point.
audit --jsonemits typed observations:That is a member-level advisory with a
kind. Adding working-tree and remote observations alongside it needs no new plane and no new gate.Proposal
Emit additional non-gating observations per member: dirty/untracked count, whether an upstream is configured, ahead/behind against it, whether member HEAD is an ancestor of the integration checkout HEAD, and the member's last commit date. Each is one
gitcall. All are advisory, exactly likeboundaryis today.The strongest single signal is the third one: a member whose commits exist on no remote ref at all is materially different from one that is merely unlanded, and nothing in Concord distinguishes them today.