Pin 13's APPEND opens "RESOLVED 2026-08-11"; actual resolution and landing were
both 2026-08-12, as the contract's own ratification line and amendment 1's both
say. The pinned date is the date the append text was drafted, frozen before the
dates diverged. Gate 10 rejected the artifact over exactly this character at
landing and annex §11.1 flagged it rather than settling it — the gate working.
The correction is appended to the ledger row; pin 13, annex §11.1 and the
original sentence are all left standing. The reason is the append-only
convention CLAUDE.md:12 states, not that history would otherwise change: 52fcf37
is immutable, and gate 10's observation remains true of it whatever a later
commit says. An earlier draft gave that false reason and §7.2 records it as
false, because it is the more persuasive one and would otherwise be re-derived.
This commit ratifies only. Per §7.0 the ledger correction is deliberately
UNSTAGED here and lands in the next commit, so that this commit is the immutable
oracle B1 reads both ORACLE and CORRECTION from. An oracle taken from the
working copy is not an oracle: an executor editing template and artifact together
would satisfy it.
B2's third part pins the ratification-to-landing surface exactly — one status
line swapped, four lines added for the marker, nothing else — so §7's frozen
body, including §7.3's template and this gate, cannot be edited during landing.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Ps1szk2mSfgp4Cz21eVH9x
Levi Neuwirth
9cc2913fab