[V7.2] Add requirement for counterparty identity verification before agent-initiated material-impact actions

Summary

Section 7.2 (Agent Authority & Session Access) covers session authorization (7.2.1), least-privilege plus human confirmation for sensitive actions (7.2.2), and log distinguishability (7.2.3). None of these require the agent to assess whether the counterparty it is acting toward — a site, service, or communication recipient — is verifiably legitimate before taking an impactful action on its behalf.

Problem

• 7.2.2 gates "sensitive actions" (payments, permission changes, data export, communications) behind human confirmation, but this is orthogonal to whether the counterparty's identity is verifiable. An agent could correctly flag an action as sensitive per policy and still act toward an impersonating or unverifiable counterparty without any additional signal being raised.
• V6.1.4 addresses browser-level detection of pages impersonating identity providers for a human user, but does not cover an agent's own counterparty-verification step when acting autonomously — for example, toward an API or a recipient with no rendered page for a human to evaluate.

Proposed addition

• BSVS-7.2.4: Verify that agentic browsing capabilities assess whether the identity of a counterparty can be verified against known legitimate identifiers before taking material-impact actions on that counterparty's behalf, independently of whether the action is already covered by 7.2.2, and that unverifiable counterparties trigger a defined fallback.

Scope clarification

This is not a duplicate of 7.2.2: 7.2.2 gates the type of action; this requirement gates the target's verifiable identity, regardless of action type. It is also not a duplicate of 6.1.4: 6.1.4 protects a human user browsing a rendered page; this covers an autonomous agent's own decision path, which may involve no rendered page at all (e.g., an agent-to-API interaction).

Suggested mapping

CWE-345 (Insufficient Verification of Data Authenticity) appears to be the closest fit — the gap is an absent verification step, not a bypass of an existing authentication mechanism. CWE-940 (Improper Verification of Source of a Communication Channel) may also apply depending on the specific interaction type. Open to maintainer input on the more precise mapping.

Happy to submit a PR with the exact requirement wording if this direction is agreed on.

Locate the repository text for Sections 7.2.2 and 6.1.4, then review the surrounding requirements and existing CWE mappings. Confirm the proposed BSVS-7.2.4 wording, fallback expectation, and preferred mapping with maintainers; done means the agreed requirement is documented consistently with the surrounding section.