provenance: identity locator fields accept truthy non-string values

[altrudev]

spec/server-provenance-v1.md defines the two identity locators as textual identifiers:

• identity.artifact.package is a Package URL;
• identity.endpoint.url is a URL.

The shared producer/consumer structural check currently tests both fields by truthiness only:

if not artifact.get("package"):
    ...

if not endpoint.get("url"):
    ...

As a result, truthy non-string JSON values such as true, 1, [1], or {"x": 1} satisfy the locator-presence check. A signed peer provenance record can therefore verify with a non-text artifact.package or endpoint.url as long as the sibling digest is valid.

This is narrower than Package URL / URI syntax validation. The immediate boundary is only the JSON primitive type the specification names: both locator fields should be non-empty strings rather than arbitrary truthy values.

Reproduction

Against current main at fc38496f3ddacaa7ce922e1218b6fcd91735d751, the exact predicates above accept each of:

true
1
[1]
{"x": 1}

Existing coverage checks a missing endpoint key digest but does not exercise non-string locator values.

Proposed bounded fix

Use one small non-empty-string boundary for artifact.package and endpoint.url in _check_structure(), exercised through both build_record() and verify_record().

Do not expand this change into full Package URL or URL syntax policy; that can be decided separately if desired.

AI-assistance disclosure: ChatGPT assisted with source triage, adversarial-case design, duplicate search, and issue drafting. altrudev reviewed the bounded claim and remains responsible for the contribution.

[github-actions]

🔴 Contributor Check: HIGH

Check	Result
Profile	HIGH
Credential	LOW
Overall	HIGH

Automated check by AgenTrust Contributor Check.