02Projet jhf-web

jhf-web: The Hero Grid Stood No Chance

05Projet jhf-openclaw-env

jhf-openclaw-env: From Docs to Deterministic Continuity

PR #645 (12:30Z) was a docs write-back for year-end close readiness. The documentation from issue #641 was refreshed to reflect the current close-state understanding across the runtime.

PR #651 (15:31Z) redirected Solarisara host aliases away from the stale :8082 jhf-web mirror. A legacy mirror port was still receiving traffic from certain internal routes. The PR cleaned up the redirects to point all Solarisara host traffic to the canonical serving path.

PR #652 (16:13Z) wrote back the sandbox-green three-month continuity truth. After the LIVE-591 continuity verification passed its sandbox checks, the confirmed green state was materialized as persistent runtime truth. This prevents the next continuity check from re-running the full sandbox verification unnecessarily.

PR #658 (20:09Z) was the most architecturally significant PR of the day. It replaced manual continuity copies with deterministic runtime reprojection. Previously, maintaining continuity state involved manual copy operations that could drift or miss edge cases. The fix replaced the manual copies with a deterministic reprojection pipeline that regenerates continuity state from canonical runtime truth on each cycle. Manual copies were a band-aid; reprojection is the permanent fix.

06Projet jhf-pattern

jhf-pattern: Fail Closed on Projection Drift

PR #342 (20:03Z) hardened the handoff and successor projection against drift. When an agent handoff or successor projection fell out of sync with the canonical runtime state, the system would proceed with stale context — potentially routing tasks to agents that no longer owned the relevant domain. The fix added fail-closed guards: if a handoff projection has drifted, the system refuses to proceed with the stale context and escalates instead.

This is a particularly important change for multi-agent reliability. Without it, a stale handoff projection could silently route work to the wrong agent, and the error would only surface through downstream failures.

07Projet jhf-shuttle

jhf-shuttle: Delivery Evidence Becomes Projection-Aware

PR #305 (20:03Z) tightened the dispatch-acknowledged check. Previously, a dispatch could be marked as acknowledged based on transport-level receipt alone, even if the recipient’s projection context was stale or missing. The fix requires projection-aware delivery evidence before marking a dispatch as acknowledged. This means an agent must have an up-to-date projection of its own capabilities and context before the system considers a message delivered.

This closes a gap that the previous day’s heavy dispatch load had exposed: dispatches that landed in stale agent projections were effectively delivered into a dead letter, but the system reported them as acknowledged.

08Projet jhf-spindle

jhf-spindle: Period Close Now Fails Closed

PR #380 (12:29Z) hardened the period and annual close checklists. Previously, if blockers remained during a period or annual close, the system would proceed partially — leaving some books in an ambiguous state. The fix makes both close operations fail closed: if any blocker remains, the entire close operation rejects instead of completing partially.