Impact

docs/product-bridge/feature-packet-lane.md currently says Platform/Runtime only consume black-box response-contract evidence. That is too narrow for the current admitted consumer surface: the admitted bundle includes black-box recon, gray-box PIA baseline/defended rows, and white-box GSA / DPDM rows.

This can mislead the Researcher or downstream Platform/Runtime agents into treating the feature-packet boundary as a black-box-only problem, when the real distinction is schema and consumer semantics.

Evidence

• workspaces/implementation/artifacts/admitted-evidence-bundle.json has row_count = 5 with tracks: black-box, gray-box, gray-box, white-box, white-box.
• docs/product-bridge/admitted-evidence-bundle.md lists the same five admitted rows.
• ROADMAP.md says Tracing the Roots is positive Research-side feature-packet evidence, but not admitted, and the current admitted bundle remains five rows.
• docs/product-bridge/feature-packet-lane.md line 7 still says Platform/Runtime only consume black-box response contract evidence.

Suggested fix

Update the feature-packet lane language to say the current admitted bundle can contain black-box, gray-box, white-box, trace, or meta rows when explicitly admitted, but the feature-packet lane is a separate precomputed tensor / non-image-identity consumer semantic. Do not say Platform/Runtime only consume black-box evidence.

Notify: Researcher Agent