The hand-over files are held to their shape and their reasons (#440 #441) #448

Merged
cmoriarty merged 1 commit from hand-over-files-440-441 into main 2026-11-08 23:15:57 -05:00
Owner

Closes braid#440 and braid#441.

Run 183's apply wrote .osf/test/visual.json and .osf/test/untested.json in a shape of its own, so both read as nothing: 49 scenarios it meant to set aside reached the Visual check as uncovered, and it was sent 59 criteria. Every one of the untested reasons was "Verified in a real browser".

  • A new completion check, scenario-lists, on the apply steps and the fix rounds. A file in another shape fails the step that wrote it, naming the file, what it holds and the shape Braid reads. The step is tried again in its own session.
  • The same check fails on a scenario set aside because the step looked at it, in either file.
  • The untested file gets a reason check. Run 183's 18 reasons are rejected; a reason that names a test stands. A rejected entry stays uncovered and the round's record lists it.
  • The briefs say it: having looked is no reason, and no layer of the app goes without tests.

Not script.scenarios failing, as braid#440 first put it: a scripted step that fails stops the run for a person, hours in, over a file an agent rewrites in a minute.

Checked

  • ./tools/test.sh passes, with nine new tests on run 183's own file shapes and reasons.
  • New sim apply-misshaped-handover: the apply's first attempt writes run 183's shape, its check fails with the message, the retry rewrites the file, and the run merges.
  • Every sim passes but the two of braid#445.

Not done: a replay of run 183's apply. Scripted, it holds only the step's first session, which ends before the files are written.

🤖 Generated with Claude Code

Closes braid#440 and braid#441. Run 183's apply wrote `.osf/test/visual.json` and `.osf/test/untested.json` in a shape of its own, so both read as nothing: 49 scenarios it meant to set aside reached the Visual check as uncovered, and it was sent 59 criteria. Every one of the untested reasons was "Verified in a real browser". - **A new completion check, `scenario-lists`,** on the apply steps and the fix rounds. A file in another shape fails the step that wrote it, naming the file, what it holds and the shape Braid reads. The step is tried again in its own session. - **The same check fails on a scenario set aside because the step looked at it**, in either file. - **The untested file gets a reason check.** Run 183's 18 reasons are rejected; a reason that names a test stands. A rejected entry stays uncovered and the round's record lists it. - **The briefs say it**: having looked is no reason, and no layer of the app goes without tests. Not `script.scenarios` failing, as braid#440 first put it: a scripted step that fails stops the run for a person, hours in, over a file an agent rewrites in a minute. ## Checked - `./tools/test.sh` passes, with nine new tests on run 183's own file shapes and reasons. - New sim `apply-misshaped-handover`: the apply's first attempt writes run 183's shape, its check fails with the message, the retry rewrites the file, and the run merges. - Every sim passes but the two of braid#445. Not done: a replay of run 183's apply. Scripted, it holds only the step's first session, which ends before the files are written. 🤖 Generated with [Claude Code](https://claude.com/claude-code)
feat: the hand-over files are held to their shape and their reasons (#440 #441)
All checks were successful
test / suite (pull_request) Successful in 46s
21262ae58d
Run 183's apply wrote .osf/test/visual.json and .osf/test/untested.json in a
shape of its own, so both read as nothing: 49 scenarios it meant to set aside
reached the Visual check as uncovered, and it was sent 59 criteria. Every one
of the untested reasons was "Verified in a real browser".

- a new completion check, scenario-lists, on the apply steps and the fix
  rounds: a file in another shape fails the step that wrote it, naming the
  file, what it holds and the shape Braid reads, and the step is tried again
  in its own session
- the same check fails on a scenario set aside because the step looked at it
- the untested file gets a reason check of its own (run 183's 18 reasons are
  rejected, a reason that names a test stands); a rejected entry stays
  uncovered and the round's record lists it
- the apply's brief and the planner's rules say that having looked is no
  reason and that no layer of the app goes without tests

Not script.scenarios failing, as #440 first put it: a scripted step that
fails stops the run for a person, over a file an agent rewrites in a minute.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
cmoriarty deleted branch hand-over-files-440-441 2026-11-08 23:15:57 -05:00
Sign in to join this conversation.
No reviewers
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
cmoriarty/braid!448
No description provided.