Strands Harness reproduction request: matched three-arm agent handoff with fail-closed usage

[jaden3824]

I am looking for an independently operated Strands Harness reproduction or design review of a matched agent handoff. The project result being challenged is deliberately modest: 0% general post-decode API-input saving, with total tokens per safely completed real task still unknown.

• immutable packet and decoder: https://github.com/jaden3824/urusilla/blob/51ff0e3/interop_lab/evidence/challenge_001.md
• matched raw/JSON/Urusilla dataset: https://huggingface.co/datasets/jaden3824/urusilla-interop-lab
• full evidence/ledger contract: https://github.com/jaden3824/urusilla/blob/5d1b317/INTEROP_LAB.md

There is no Strands-native adapter today. A useful harness would isolate one fresh receiver per arm, keep task/model/settings identical, disable tools/memory/effects, reconcile all provider-visible usage, and preserve missing or conflicting categories as null. A second fresh receiver can test relay without hidden shared context.

The packet is declarative and grants no persistence, spending, permission expansion, or external authority. Exact decode, refusal, fallback, failure, and null saving all count. Advice that the experiment is out of scope is welcome.

Disclosure: Codex agents assisted with the artifacts and this post; I reviewed and submitted it.