You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
The implementing rounds' bound is a brake, and moves into kernel/brakes.py before the author Role is named #874
A follow-on to "While the draft the Factory opened fails its checks, the issue admits another implementing Run" (#827, record 0217). Codex found it on #827, and by the Factory Builder's ruling it lands in its own pull request so that #827's approval stands.
What is wrong
kernel/AGENTS.md says a new brake goes in kernel/brakes.py, the protected path. Record 0217's bound on implementing rounds is written directly in kernel/issue.py's _implementing: at max_attempts completed implementing Runs since a writer acted, the issue is held attempts_exhausted.
It counts completed Runs whatever they concluded. brakes.run_ceilings counts failed attempts, so it does not hold this bound on its own.
The bound works and the rule is dormant, since config/defaults.json names no author Role. So nothing fails open. What is wrong is only where the brake lives: a change to it is not seen as a change to a protected path.
What is asked
The implementing rounds' bound moves into kernel/brakes.py, as a function kernel/issue.py asks with a Run in flight or none, giving the same Hold and evidence.
Record 0217's boundary row names the new site and its mutants.
Activity