Task · CXO-0032

Adopt the 2026-09-05 wire additions: turn_trigger, safety-buffering reasons, reasoning context, access programs, and the unnamed keepalive and response.metadata events

Description

A full unsampled drift scan on 2026-09-05 (70 archives, 9.4 GB compressed, 25.0 GB decompressed, 9,962,295 lines, 0 undecodable) against the 2026-09-03 baseline produced 0 breaking and 215 new findings. The baseline was refreshed from that scan the same day (Rob authorised it), so the live drift gauge no longer reports these; this task carries the shapes that are worth adopting or naming so they are not lost in the refresh.

Shape additions worth adopting (all content-free field names; values are protocol enums, never conversation text):

Cardinality checks the same scan makes necessary:

Acceptance Criteria

Definition of Done

Implementation Plan

Wave 2: commit root-owned seams first; frozen owned lanes implement with synthetic tests; integrate, just check, one corpus confidence gate and CodeRabbit; one push; watchtower-only deploy observation and m7kni proof; reconcile acceptance by evidence layer.

Implementation Notes

Clean main e8e97fd directly descends from c567894 and equals origin/main. CI 33988760737 and release-please 33988761017 succeeded. D8 holds; traces and agento11y remain disabled.

Final Summary

Wave 2 delivered at aca5e5de0bcd4ba6f5fd72dfaa6caef3a6c7fb71. Full content-free inventory: 70 files, 9962295 lines; one trigger enum, 11 tool names (cap64 retained), models Observed updated. Decoder names keepalive/response.metadata. Full unsampled baseline refresh: 21 named events, zero undecodable, baseline byte-identical; signature test and sampled probe pass. access_programs.cyber omitted because frozen Turn seam lacks a field; namespace deferred. just check passed at that source SHA; CI 33993480890 success. Watchtower deployed that SHA healthy, restart count 0. Publish 33993481221 failed signing after successful manifest push; run-level publication completion remains open. The single D20 corpus gate was canceled (exit 143), so corpus confidence is not proven. Final tracker closeout is local under the one-push contract.

References

View the source file on GitHub