fix(agents): demote abandoned ACTIVE rows so agent-record can pass - #177
fix(agents): demote abandoned ACTIVE rows so agent-record can pass#177bakon11 wants to merge 1 commit into
Conversation
c0c0120 to
9bc6fc4
Compare
Live-row audit: ACTIVE rows with no Git evidence → SPIKE so agent-record passes. Re-pin RUNNABLE_BASELINE + unit test. Document leftover Gemma4/ROCm lab env knobs and server prompt/max_tokens caps. Roll non-canonical BENCHMARKS H2 into benchmark-record; keep STATUS under ratchet. FOLLOWING_AGENTS_PROTOCOL Following-Agents-Protocol: true AI-Assisted: true Assisted-by: Hermes:grok-4.5 [Hermes]
9bc6fc4 to
9caccbc
Compare
|
Thanks for this — it was a real red gate and the diagnosis was right at the time. Closing it as superseded, with the evidence, because the situation changed underneath it. The premise no longer holds. #177 exists because
And merging it now would be actively wrong, not merely redundant. Three of the rows this PR demotes
The Nothing here is a criticism of the change — the rows genuinely looked abandoned when you audited them, and the branch-and-main-commit heuristic is the right one. It is just that the underlying drift got fixed from the other direction first. If you want to keep the thread going, the durable version of this problem is still open: |
Why
agent-record/test_audit_live_rows.py::test_shipped_record_has_no_abandoned_active_rowfails on cleanmain. That reds every PR (including #163 and #154).Supersedes closed #173 (head tracking stuck on an older SHA).
What
ACTIVE→SPIKEwhere audit reports no branch and no main commit mentioning the row IDengine-matrix.mdlifecycle Total SPIKE/ACTIVE countsVerify (local green)
Unblocks merge CI for the rest of the queue.
FOLLOWING_AGENTS_PROTOCOL
Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Hermes:grok-4.5 [Hermes]