Runtime Invariants and Limits¶

This page states the public ground rules of the repository: what must stay true even while weaker evidence families are still being recovered, and what should not be inferred from the current outputs.

It is deliberately short because these are the rules that should remain clear even when the rest of the handbook becomes more detailed.

Invariants¶

commands rewrite governed roots rather than ad hoc destinations
source-family ownership stays visible in tracked data layout
publication outputs remain downstream of tracked repository state
docs and review surfaces must not hide weak evidence posture

These invariants are not implementation preference. They are the conditions that keep the public product reviewable.

What These Invariants Mean In Practice¶

a visible output should always trace back to a stable repository-owned path
one source family should not disappear into a generic mixed bucket
maps and reports should remain the downstream explanation, not the upstream proof
caveats and refusal language should survive even when the interface becomes cleaner

Definition Of Done¶

A change is not done when files merely exist. It is done when the changed boundary is reviewable, linked, and validated at the right layer.

For the public product, that means:

the changed surface can be inspected
the next narrower explanation still exists
the wording still matches the available proof
the repository has not silently traded clarity for elegance

Dependency Governance¶

Prefer explicit runtime contracts and stable checked-in files over hidden side effects or opaque rebuild steps.

That matters because people should not need maintainer folklore to understand how a claim became visible.

Known Limits¶

animal aDNA remains a partial recovery surface
atlas presence is not the same as scientific completeness
docs breadth can regress if it is not tested explicitly

What The Current Outputs Still Cannot Honestly Claim¶

that the animal aDNA side is already a finished, evenly recovered evidence engine
that all visible points carry the same locality and chronology strength
that broader geography automatically means deeper proof
that a cleaner handbook means the repository itself has become more complete

Risk Posture¶

The main risk is elegant-looking narrowing: fewer pages, fewer links, and fewer explicit caveats while the repository still claims to explain the same evidence landscape.

That is why the repository treats clarity as part of technical quality. A surface that sounds cleaner but says less is often a regression, not an improvement.