Abyss Vaults: Low in the historical state

Question and review date

The review assessed Abyss Vaults reward accounting under conditions documented on 9 October 2026. It connected the Move package lineage, historical events and operation at that time.

What the model evidence establishes

The isolated check established the examined relationship in the model. It did not show that a regular user could create the required operational condition and carry out a persistent attack. That unresolved prerequisite determined the rating.

Balance check and countercheck

The artificial balance below shows a basic control: the expected result should pass and the deliberately differing closing balance should fail. It illustrates a test property, not Abyss code or production loss.

1def check_balance(opening, incoming, outgoing, closing):
2    values = (opening, incoming, outgoing, closing)
3    assert all(type(v) is int and v >= 0 for v in values)
4    return opening + incoming - outgoing == closing
5
6assert check_balance(100, 20, 15, 105)
7assert not check_balance(100, 20, 15, 106)

Version comparison

The relevant module content was unchanged within the examined package lineage. Package names alone would therefore not establish different behaviour. The hash block illustrates byte comparison using invented inputs.

1import hashlib
2
3def artifact_id(content):
4    return hashlib.sha256(content).hexdigest()
5
6assert artifact_id(b"model-v1") == artifact_id(b"model-v1")
7assert artifact_id(b"model-v1") != artifact_id(b"model-v2")

History and activity at sampling time

Historical events did not show that the same operational condition regularly existed at sampling time. Not every observed difference could be uniquely attributed to one cause. The model therefore separates isolated evidence from a production claim.

1record = {"isolated_check": True, "production_reachability": False}
2
3def supports_production_claim(evidence):
4    return evidence["isolated_check"] and evidence["production_reachability"]
5
6assert not supports_production_claim(record)

Alternatives and rating

Alternative candidates were assessed separately, either rejected or classified as limited hardening questions. The updated assessment remained Low for the historical state; Medium applied only under changed economic conditions. High and Critical were not established.

What reassessment requires

Reassessment requires fresh package and operational evidence. These results date to October 2026; this report includes no fresh state review.

Review limits

• As of 9 October 2026. Model evidence and operational reach are assessed separately.

References

The links provide project context. The judgment rests on the evidence described in the audit. Review models and actual tool excerpts are labelled separately.