# Decoder and Harness: editorial review

Reviewed September 5, 2026. The original publication dates remain August 22 for
Decoder and February 18 for Harness.

Decoder presents architecture guidance and illustrative pseudocode alongside a
[dated local ZJX source note](./zjx-infrastructure-hardening-source-note-2026-08-22.md).
That note does not verify every recommendation. This correction reports no new
decoder experiment or general security certification.

Harness combines selected research with Greyforge operating proposals. The public sample now identifies its diagnostic ordering as a proposal. Earlier
wording overstated support for comparisons between harness changes and model
upgrades. The edition reports no controlled Greyforge reliability, security,
cost or model-swap result. Its taxonomy and scorecard are proposed frameworks;
the public trace is illustrative, not a measured run.

The primary research supports scoped findings about repository tasks,
interfaces, interactive evaluation and action-observation systems:
[SWE-bench](https://arxiv.org/abs/2310.06770),
[SWE-agent](https://arxiv.org/abs/2405.15793),
[AgentBench](https://arxiv.org/abs/2308.03688),
[OSWorld](https://arxiv.org/abs/2404.07972),
[AgentDojo](https://arxiv.org/abs/2406.13352) and
[ReAct](https://arxiv.org/abs/2210.03629).
These papers do not independently validate the proposed Greyforge framework.
This is a selected literature synthesis, not a systematic review or a current
model ranking.

Systems references checked for the Decoder correction include the official
[Python subprocess documentation](https://docs.python.org/3/library/subprocess.html#subprocess.Popen.communicate),
[Python archive extraction guidance](https://docs.python.org/3/library/tarfile.html#extraction-filters)
and [XZ manual](https://tukaani.org/xz/man/xz.1.html).
Implementation and platform details still require their own review.