Batch 41.3 Group 2 Benchmark Alignment
This page summarizes the May 21, 2026 rerun of 10 Tier 2 global skills after fixture route remediation. It links the execution outcome back to the reports, generated benchmark data, validation, and next triage route.
10 global skills, both agents, 3 runs per agent where not infrastructure-blocked.
7 skills improved above 0% for at least one runner; 3 remain priority triage targets.
Verify gates, benchmark runs, generated-data freshness, benchmark coverage, routing test, and whitespace checks passed.
Skill Results
| Skill | Claude | Codex | Follow-up |
|---|---|---|---|
dogfood | 0/3 | 3/3 | Claude owner-scenario/manual-check assertions. |
expert-review | 3/3 | 3/3 | No immediate triage. |
guide | 2/3 | 3/3 | One Claude artifact assertion miss. |
handoff | 0/3 | 3/3 | Claude handoff artifact expectations. |
hygiene | 0/3 | 1/1, 2 blocked | Claude artifact expectations and Codex timeouts. |
migrate | 0/3 | 0/0, 3 blocked | Priority triage: migration-plan artifact and runner timeout. |
mono-plan | 0/3 | 1/2, 1 blocked | Claude artifact expectations and Codex partial pass. |
pack | 0/3 | 3/3 | Claude artifact expectations. |
prototype | 0/3 | 0/1, 2 blocked | Priority triage: hub/clickable/fake-data expectations. |
provision-agentic-config | 0/3 | 0/3 | Priority triage: orchestration and monorepo safety assertions. |
Shipping Boundary
The committed boundary should include the 10 report updates, generated
benchmark/showcase assets, the routing-graph performance fix, and
task/history notes. The next recommended command is
$session-triage provision-agentic-config benchmark failure.