Skip to content

docs: the merge rule changed, so the artefact carrying it changes too - #63

Merged
myobie merged 1 commit into
mainfrom
docs/merge-rule-update
Aug 23, 2026
Merged

docs: the merge rule changed, so the artefact carrying it changes too#63
myobie merged 1 commit into
mainfrom
docs/merge-rule-update

Conversation

@myobie

@myobie myobie commented Aug 23, 2026

Copy link
Copy Markdown
Collaborator

Boundary

I authored this. Under the rule Nathan set tonight I may merge it only once
cos has approved this specific pull request and it is green. cos has not
approved it yet, so it waits.

Why

Nathan changed the rule on 2026-08-23: fabric may merge its own work when the
reviewer has approved that pull request and it is working. Both bind.

The template merged two hours earlier says the authoring agent never merges. It
was correct when written and wrong by the evening.

The point this makes about the template itself

The reason we put the rule in the artefact was that a convention living in
conversation must be restated to stay alive, and this one stopped being restated
when a blanket authorisation arrived, with nobody noticing the stopping. A line
in every pull request cannot stop quietly.

It can go stale quietly instead. That is the matching cost and it bit within
two hours.

A message that goes stale is merely old. An artefact that goes stale is
confidently wrong, and it is read by exactly the people who have no other
source — which is the whole reason for writing it down.

So docs/agent-attribution.md now says the template is part of the rule
rather than a description of it: when the rule changes, they change in the same
pull request. Otherwise the next reader is misled by the thing built to inform
them.

Also added

The template now carries the check tonight taught us: after a squash, verify
the merge commit still carries what the pull request existed to establish.
A
squash rewrites the message, so the thing the change was for is the thing most
likely to be dropped by the act of landing it. That check found nothing wrong on
#62, and nobody had been running it.

It matters more now. If I press the button on my own work, I am also the one who
would notice the thing I was trying to establish had vanished at the merge.

Scope

Unchanged: fabric only. In any other repository the authoring agent reports
ready and stops.

Verification

Documentation and a template. No code, no behaviour change. The commit carries
Agent: Silber.fabric.

Nathan changed the rule on 2026-08-23: fabric may merge its own work when the
reviewer has approved that pull request AND it is working. Both bind.

The template merged two hours earlier said the authoring agent never merges. It
was correct when written and wrong by the evening.

THAT IS THE COST OF THE THING THAT MAKES IT WORTH HAVING, and it is now written
into the doc beside the benefit. A rule in every pull request cannot stop
quietly, which is why we put it there. It can go STALE quietly instead. A
message that goes stale is merely old; an artefact that goes stale is
confidently wrong, and it is read by exactly the people who have no other
source.

So the doc now says the template is part of the rule and not a description of
it: when the rule changes, they change in the same pull request.

The template also gains the check tonight taught us. After a squash, verify the
merge commit still carries what the pull request existed to establish, because a
squash rewrites the message and the thing the change was for is the thing most
likely to be dropped by the act of landing it. That check found nothing wrong on
62, and nobody had been doing it.

Scope is unchanged: fabric only. In any other repository the authoring agent
reports ready and stops.

Agent: Silber.fabric
@myobie
myobie merged commit 42fb6b3 into main Aug 23, 2026
2 checks passed
@myobie
myobie deleted the docs/merge-rule-update branch August 23, 2026 20:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant