Issues

• Instant recall's confirm and verify steps are not gated by the spend ceiling

  Smana/runlore#589 · Filter by author Smana opened on Sep 26, 2026

• Model-comparison report publishes vacuous rubric, coverage and confident-wrong columns when cases lack ground_truth

  Smana/runlore#394 · Filter by author Smana opened on Aug 2, 2026

• Gemini 3.x unusable: client omits thought_signature on function-call parts, investigations fail on turn 2

  Smana/runlore#392 · Filter by author Smana opened on Aug 2, 2026

• glm-4.6 stalls indefinitely on forced tool_choice, silently degrading verify, recall reranking and the eval judge

  Smana/runlore#391 · Filter by author Smana opened on Aug 2, 2026

• Security: wrap untrusted tool outputs in an explicit delimiter (project-wide)

  Smana/runlore#362 · Filter by author aqemia-smaine-kahlouch opened on Jul 23, 2026