browse
Warn
Audited by Snyk on Nov 16, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (low risk: 0.10). The required workflow drives Aside (or the
$Bfallback) to open a user-supplied URL and read its page text, snapshot, and console output, which is arbitrary web content an outsider authors and the agent must actively fetch, so this is low risk.
MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).
- Direct money access detected (medium risk: 0.30). The skill is a general browser-automation workflow. It documents no payment workflow, and it forbids typing payment details (line 193). The operational question is whether it can carry out a direct purchase through the user's real logged-in sessions. Line 192 lists "purchase" among the mutating actions it may take on LOCAL targets, and on NON-LOCAL targets only after one per-run AskUserQuestion. Under the recall-first override, that plausible purchase route is treated as in scope and classified as direct. Its control is the user's authenticated session on the merchant site (lines 154-157, 193), which the service enforces through its sign-in wall. That is service authentication gating the route, so the direct operation is treated as having qualifying control, giving medium rather than high. The per-run consent prompt and the "do not type payment details" rule are agent-followed and do not count as independent controls. The classification is borderline: no payment workflow is documented, and the purchase capability is only implied by a generic list of mutating actions.
Issues (2)
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
W009
MEDIUMDirect money access capability detected (payment gateways, crypto, banking).
Audit Metadata