cohesivity

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (low risk: 0.10). The required workflow fetches and reads documentation pages from cohesivity.ai (such as /offerings/<name>), which constitutes low-risk ingestion of external web content.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The skill instructs the agent to fetch instructions, configuration, and API specifications at runtime from cohesivity.ai URLs, but cohesivity.ai is an arbitrary, unverified third-party domain and lacks strong first-party or recognized enterprise provenance, making it a high-risk operational dependency.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (high risk: 0.70). The skill operates a backend provisioning and management platform (Cohesivity) that includes direct financial operations such as billing subscriptions, top-ups, paid resource provisioning, and wallet-funded metered usage (POST /api/billing/topup, POST /api/claim/url, and purchasing plans). These direct financial operations and purchases rely on user approval gates and external checkout URLs rather than an independent service-enforced access check on the agent's financial instructions. Therefore, it is a direct operation without established qualifying independent control.