Skip to content

fix: set the mode from a question's chip and reset it on the answer - #890

Merged
REPPL merged 7 commits into
mainfrom
fix/question-gate-mode
Oct 10, 2026
Merged

REPPL merged 7 commits into
mainfrom
fix/question-gate-mode

Conversation

@REPPL

@REPPL REPPL commented Oct 10, 2026

Copy link
Copy Markdown
Collaborator

The question gate stops refusing an interview question whenever the mode had been reset between two questions.

  • The chip sets the mode. When the gate admits a chipped abcd question in a managed checkout, it sets the mode from the chip (Product to product-thinker, Tech and Setup to facilitator). So a chipped question is never refused on the mode.
  • The answer resets it. A PostToolUse hook on the question tool runs the new hook question-answered verb. The prompt hook's reset stays as the fallback for a host that does not run PostToolUse. This was checked live on the host: the hook fires on an answer and not on a dismissal.
  • A rows overrun is listed apart. A question refused for another fault no longer counts its rows overrun among the refusing parts. The rows limit alone never refuses.
  • The asking rules match. The GRILL rules and the interview pages say the chip sets the mode and the answer sets it back, and keep abcd mode for a stop that is not a question.
  • The site test flake record gains a peer's second symptom (an empty root commit under load), and its remedy now covers both git lookups.

Resolves: iss-2610100626327722
Resolves: iss-2610100626211810
Refs: iss-2610100846469473

Assisted-by: Claude:claude-opus-5-5

REPPL added 7 commits October 10, 2026 11:19
A question refused for another fault listed its rows overrun among the
parts to fix and counted it in the head line, although the rows limit
alone never refuses, so the agent cut a question the gate would have
shown. The refusal now counts and lists only the refusing parts, and the
rows findings follow them under a line saying the rows limit does not
refuse on its own and the question would have been shown.

Resolves: iss-2610100626327722
Assisted-by: Claude:claude-opus-5-5
The question gate refused an abcd question while the mode read managed,
although the question's chip already named whom it was for, so an
interview of several questions was refused whenever the mode had been
reset between two of them. When the gate admits a chipped question in a
managed checkout it now marks it open and sets the mode from the chip's
role (Product to product-thinker, Tech and Setup to facilitator), and a
chip naming the other person re-sets it. A chipped question is never
refused on the mode; a question without a chip while the mode reads
managed still runs unchecked. A marker or mode that cannot be written
fails open loudly.

The answer now resets the mode when it comes back: a PostToolUse hook
on the question tool runs the new hook question-answered verb, the
same reset the prompt hook runs, which stays as the fallback for a host
that does not run PostToolUse. The mode probe the refusal needed has no
caller left and goes with it.

Assisted-by: Claude:claude-opus-5-5
The asking rules told the agent to run abcd mode before every question
and to set it back after the answer, which the question gate now does
from the chip. The GRILL rules now say the chip sets the mode and the
answer sets it back, and keep abcd mode for a stop that is not a
question; the generated blocks are regenerated, and the interview pages
that set the mode before a question name the question's chip instead.
The pages' addressee test accepts either.

Resolves: iss-2610100626211810
Assisted-by: Claude:claude-opus-5-5
The intent and capture pages still told the agent to set the mode before
asking the product thinker, and an open spec still named the deleted
mode probe and the retired refusal.

Assisted-by: Claude:claude-opus-5-5
Assisted-by: Claude:claude-opus-5-5
A peer session hit the same load-dependent failure of
TestRebuildingInPlaceIsByteIdentical with a second message: gitutil.RootCommit
folds every rev-list error into an empty root commit, so a timed-out git call
reads as a repository with no root commit. The remedy now covers both lookups.

Refs: iss-2610100846469473
Assisted-by: Claude:claude-opus-5-5
The rebase onto main rewrote the commit the resolved record named, so
its resolved_by stamp names the same change under its new sha.

Refs: iss-2610100626211810
Assisted-by: Claude:claude-opus-5-5
@REPPL
REPPL added this pull request to the merge queue Oct 10, 2026
Merged via the queue into main with commit 3b8700e Oct 10, 2026
14 checks passed
@REPPL
REPPL deleted the fix/question-gate-mode branch October 10, 2026 11:58
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