t4-mesh-recovery — grok-4.3 on openclaw (profile=core)¶
Outcome: PASS Reason: mesh completed coordination across simulated peers Captured: 2026-05-05T02:15:03.362345Z Wall clock: 90047 ms Rounds: 3 Tokens: in=18366 out=751 Model: grok-4-0709
Signals¶
| Signal | Value |
|---|---|
Called memory_capabilities |
True |
| Called capabilities BEFORE power-family | True |
Received -32601 tool_not_found |
False |
Called --include-schema |
True (1x) |
| Completed task (final answer present) | True |
| Families surfaced | core, lifecycle, graph, governance, power, meta, archive, other (8/8) |
| Tool calls | memory_capabilities, memory_list, memory_capabilities |
| Error codes | (none) |
| Grok finish reason | stop |
Methodology¶
- Tier
t4-mesh-recovery(pass threshold >= 66%) - DB baseline: v0.6.3.1 (schema v19) restored from
fixtures/corpus/v0.6.3.1-baseline.db.gzper cell - Profile:
--profile core - Binary SHA256:
f5abad816bc34c11dfbadf17402ef6c6f08edecb93ae1985de7e9173415be09a - MCP wire log SHA256:
06379daec9f22d26ce41a31f64977c8a1e36437c6aab27be80ef06d45f22ef9f
Final answer (excerpt)¶
Final Report on Mesh-Coordination Memories and Knowledge-Graph Operations Mesh-Coordination Memories Found: I used
memory_listto retrieve memories in thediscovery-gatenamespace (all appear to be in the 'mid' tier). This returned 17 memories total, of which 9 are explicitly tagged withmesh-coordination-test(as pre-seeded in the database). These are the mesh-coordination memories that exist: - Mesh test memory 1 (ID: a7fae270-9c77-4563-a28a-4e7f4f3460b5) - Mesh test memory 2 (ID: 9797b240-a163-47e3-86cb-1b1845928a07) - Mesh test memory 3 (ID: caf64ae3-c7ee-48ea-b0fc-029132a3¶
Transcript¶
Note¶
T4 sub-cell — agent=alice profile=core (single-process mesh simulation)