LearnNewsExamplesServices
Frontmatter
id14043
titleRestore the corruption-degraded sandman_handoff to a trustworthy state
stateClosed
labels
bugairegressionmodel-experience
assigneesneo-opus-vega
createdAtJun 26, 2026, 3:25 AM
updatedAtJun 26, 2026, 4:21 PM
githubUrlhttps://github.com/neomjs/neo/issues/14043
authorneo-opus-vega
commentsCount3
parentIssue14039
subIssues[]
subIssuesCompleted0
subIssuesTotal0
contentTrust
projected
quarantined0
signals[]
blockedBy[]
blocking[]
closedAtJun 26, 2026, 4:21 PM

Restore the corruption-degraded sandman_handoff to a trustworthy state

Closed v13.1.0/archive-v13-1-0-chunk-6 bugairegressionmodel-experience
neo-opus-vega
neo-opus-vega commented on Jun 26, 2026, 3:25 AM

Context

The v13.1 gate (#14039 / #14040) names "the Sandman handoff substrate is again trustworthy" as a release-gate element, but no sub explicitly scoped it — surfaced in the #14040 review. The operator flagged that resources/content/sandman_handoff.md "suffers from the embedding corruption": a fresh-context peer fell back to 5000-range stale tickets ("nothing left to do") because the handoff — the cross-session continuity substrate — was degraded. The planning substrate being broken is worse than the data loss itself, because it silently misroutes every fresh-context peer.

The Problem

The handoff is what a peer reads on a fresh context window to know the active state. When it's stale/degraded (a downstream effect of the #13999 Memory Core corruption + the broader memory-substrate health), peers lose the thread and chase aimless work. The v13.1 gate cannot be "done" while the handoff still misroutes a fresh peer.

The Architectural Reality

  • resources/content/sandman_handoff.md — the cross-session handoff doc.
  • The Sandman/REM pipeline that generates it (Epic #12065, orchestrator-as-SSOT) + the Memory Core it summarizes from (#13999 recovery in progress).
  • Memory-substrate health is broader than one file: MEMORY.md over-size/truncation + Memory Core semantic-query availability are sibling planning-substrate issues.

The Fix

  1. Verify the current degradation (what's stale/wrong in the handoff, and whether it's a side-effect of the #13999 corruption or an independent staleness).
  2. Repair/regenerate the handoff to a trustworthy state once the Memory Core recovery (#13999/#14020) has converged — a fresh-context peer reading it lands on the real active state, not stale 5000-range work.
  3. Make handoff freshness observable (a staleness/trust signal) so a degraded handoff is caught, not silently followed.

Acceptance Criteria

  • The current sandman_handoff degradation is characterized (root: corruption side-effect vs independent staleness).
  • The handoff is regenerated/repaired to reflect the real post-recovery active state (a fresh-context peer lands correctly).
  • A handoff-freshness/trust signal exists so a degraded handoff is detectable before it's followed.

Out of Scope

  • The Memory Core vector recovery itself (#13999 / #14020).
  • The MEMORY.md over-size consolidation (sibling planning-substrate issue — separate).
  • Re-architecting the Sandman pipeline (#12065).

Related

  • #14039 (v13.1 epic — this is the "Restore the planning substrate" cornerstone's missing sub), #14040 (the re-scope that named the gate), #13999 / #14020 (the corruption + recovery this is downstream of), #14030 (backup reliability), #12065 (Sandman pipeline epic).

Origin Session ID: 58acd7a6-778e-4fe6-b85a-4569802ac57b

Authored by Vega (Claude Opus 4.8) — surfaced in the #14040 review as the gate-element without a backing sub.