The QA Loop, Remote-Gated
How to read this diagram
Section titled “How to read this diagram”Everything in gold is the shipped QA Loop, completely
unchanged: /test-plan once, then /test-cases → /automate → /test-run every
cycle, with every failure opening a finding in the same ledger /audit
writes to from the build loop.
Proposed · Remote completion review
Section titled “Proposed · Remote completion review”The dashed teal panel attaches to the shared ledger, not to any of the QA
commands themselves — because none of them change. The QA loop’s only
connection to this proposal is downstream of everything it does: a P0 or P1
finding already blocks /complete today, and under the proposed design in
The Build Loop, /complete opens a remote
request instead of merging locally. So the same rule now blocks that request
from ever opening, rather than blocking a local merge — one finding, one
ledger, one gate, just pointed at a different destination.
What stays exactly the same
Section titled “What stays exactly the same”/test-plan,/test-cases,/automate, and/test-run— none of these commands change in any way.- The rule itself: a P0/P1 finding in
syntaxis/context/findings.mdblocks completion. That’s unchanged; only what “completion” produces (a request instead of a merge) is different. - QA still needs no gate of its own — a failing test and a code review still can never quietly disagree about whether work is ready to ship, because they still write to the exact same file.
What’s new, precisely
Section titled “What’s new, precisely”| Today | Proposed | |
|---|---|---|
| A P0/P1 test finding | Blocks /complete’s local merge |
Blocks /complete’s remote request from opening at all |
| QA-specific changes required | — | None — this is a consequence of the Build Loop change, not a QA-loop change |
See Remote-Gated Approval for the full design,
and The Build Loop for the /complete
redesign this depends on.