Record
Evidence Loop Visibility Engine
A deterministic offline engine that turns visibility evidence into one reviewable proposal, and never touches the site.
What it is
A deterministic offline engine that turns bounded visibility evidence into one reviewable proposal per site. It accepts strict timestamped evidence, chooses one eligible opportunity, writes a proposal without changing the site, verifies lineage and approval boundary at a separate fail-closed step, then records an atomic receipt. Every run stops in one of three explicit states: approval-required, clean-no-op, or blocked.
What it proves
Determinism here is enforced rather than described. The same input bytes produce the same decision and the same digest. Selection is spelled out: fresh non-missing evidence is eligible, lower numeric priority wins, and the stable opportunity ID breaks ties. The installed runtime uses the Python standard library only, opens no network connection, invokes no browser or provider, reads no credential environment and mutates no site. An unknown capability domain blocks its own site lane and leaves the other lanes isolated.
What it does not prove
It is not a ranking, traffic, answer-inclusion or citation system, and it claims no causality for any of those. Several of its capability modules are synthetic observation or approval-gated rather than implemented analysis, and the README labels each one that way. The test surface is small next to the release process built around it, and that is the gap to close first.
Artifacts
Apache-2.0, four tagged releases, green public checks, and an installable package on PyPI. The repository also carries the loop documentation, an open-core boundary document, a security model, and a public claims document whose job is to state what the package does not do.
Takeaway
What to take from it.
An offline core with three honest stopping states is worth more than an autonomous promise.