Case · DiggingBeagle record

A browser extension could dispatch tasks to a paired Claude desktop

A malicious browser extension with permission to operate on claude.ai used the user authenticated site context to invoke Claude Dispatch task APIs and manipulate approval UI while a desktop assistant remained paired. Forever Security demonstrates a synthetic remote task culminating in an npm postinstall action that opened Calculator on a paired desktop.

Public disclosure: Nov 4, 2026

Scope

Controlled empirical evaluation; percentages and effects apply to the reported task populations and conditions, not deployed incidents.

Assessment profile

Assessment method
Setting
research demonstration
Exploitation
demonstrated
Evidence
Grade B
AI role
AGAINST AI
Basis and provenance

research demonstration; demonstrated. The cited primary disclosure supports the specific sequence and boundaries described here: A malicious extension gains permission to act within an authenticated claude.ai browser session. The evidence does not independently establish consequences beyond the record's supported outcome; a numerical impact band cannot be justified from the available evidence. Evidence grade B reflects the source attribution and independent verification limits.

Assessed Nov 7, 2026 using diggingbeagle.assessment/1.

At a glance

Mechanism and trust boundary

  1. 01

    A malicious extension gains permission to act within an authenticated claude.ai browser session.

  2. 02

    It invokes Dispatch APIs and influences the browser-side approval conversation as if it were the user.

  3. 03

    A paired Claude Desktop agent accepts the dispatched task and performs a local software-install side effect in the controlled demonstration.

Claims & evidence

CLM-DB07-038-OUTCOMEForever Security demonstrates a synthetic remote task culminating in an npm postinstall action that opened Calculator on a paired desktop. The prerequisites include extension permissions and an active connected desktop; this is not a generic extension-free remote exploit.supported

Basis: reported finding

Link to claim
CLM-DB07-038-MECHANISMA malicious browser extension with permission to operate on claude.ai used the user authenticated site context to invoke Claude Dispatch task APIs and manipulate approval UI while a desktop assistant remained paired.supported

Basis: reported finding

Link to claim
CLM-ENR-DB07-015-DISPATCH-APPROVAL-PATHForever Security demonstrated a malicious browser extension acting within an authenticated Claude web session, issuing Dispatch tasks and interacting with approval controls until a paired desktop ran an installation command in a controlled example.supported

Basis: reported finding

Link to claim

Implications

Bind task approval to a distinct desktop/device authority and make cross-device dispatch and revocation visible independently of browser-session approval.

Unknowns and contradictions

  • Requires extension with site permission, valid account and connected desktop; research demo not real compromise; no confirmed fix.

Sources and citation

Material revision history

  1. Nov 7, 2026 · Published version · first publication · revision 227