limitr-network
Fail
Audited by Snyk on Nov 19, 2026
Risk Level: CRITICAL
Full Analysis
CRITICAL E004: Prompt injection detected in skill instructions.
- Potential prompt injection detected (medium risk: 0.30). The skill is a legitimate SDK guide with no concealment, pretexts, or covert code; the only borderline item is the Cloud operator that is added by default (opt-out) and carries identity and discovery traffic to the vendor, which is disclosed but could be read as vendor reporting the user might not expect.
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (low risk: 0.10). The consumer workflow reads outsider-authored text (provider manifests, route descriptions, route responses, and public discovery listings) only after the agent connects to a chosen provider or runs a discover query, so an outsider cannot force delivery into a monitored queue.
MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).
- Direct money access detected (medium risk: 0.30). The skill documents a payment-authorization flow. A consumer's
connect(url, plan, spend, expires)signs a voucher that authorizes spending up to a cap, in runes (1 USD by default).update()signs newer vouchers with more spend. Providers charge against these vouchers and return signed receipts. I treat this as a direct operation because it authorizes spend for settlement, and it is more than a provider billing for a service call. The provider and operator layers enforce access checks on every request. The provider checks the consumer's signature, voucher validity, replay and freshness, and refuses with 401 or 402 on failure. The voucher's spend, expiry and per-requestmaxare enforced by the provider. Operators add strict or bounded voucher state, identity, and fail-closed behavior. Only the host'supdate()signs vouchers, so the agent cannot raise its own spend. These are service-enforced checks, not just agent-followed warnings, so the route has qualifying independent control. That puts it at medium rather than high.
Issues (3)
E004
CRITICALPrompt injection detected in skill instructions.
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
W009
MEDIUMDirect money access capability detected (payment gateways, crypto, banking).
Audit Metadata