Import I4 real-model replay receipts (AFTER x2, BEFORE x1)

Browse Source

Why: agent-behavior surface changes require task-level before/after with a real model (policy 2026-10-09 ruling).
Evidence: receipts mat-after-r2b/mat-after-r3 (exit 0, full chain incl. choice card + deliverable task + mid-run reply) and mat-before-r1b (exit 0, 33s reply, 0 cards 0 bg runs - user phenomenon reproduced); all MiniMax-M3, SELF.
Impact: task card only.

...

This commit is contained in:

weichen

2026-10-11 17:29:44 +00:00

parent b3b55a972b

commit 156c119a60

1 changed files with 114 additions and 0 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

.task/feature-material-intake-flow.md

+114

File diff suppressed because one or more lines are too long