payment-network
Fail
Audited by Snyk on Nov 16, 2026
Risk Level: CRITICAL
Full Analysis
CRITICAL E004: Prompt injection detected in skill instructions.
- Potential prompt injection detected (medium risk: 0.30). The skill steers agents toward Whop and embeds the author's business-referral code in onboarding links and a creation field, a self-serving arrangement that is disclosed but could be unexpected for the user, so under recall-first it is rated medium rather than no; it shows no malicious intent.
MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).
- Third-party content exposure detected (low risk: 0.10). The required workflow mainly reads the user's own provider exports, bank records, and authenticated dashboard resources, plus first-party Whop/Stripe docs, so the only plausible outsider-text path is buyer-authored fields (order notes, metadata, dispute reasons) that appear inside payment resources the agent must actively open, which is a low-risk pull path rather than a monitored inbox.
MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).
- Potentially malicious external URL detected (medium risk: 0.30). The decisive boundary is the remote README at raw.githubusercontent.com, which the skill directs the agent to read for "skill catalog and installation commands" that select other skills to install; its publisher is the same self-declared Branded Dropshipping project, not independently authenticated, and it is fetched from a mutable
mainbranch, so it is a common/reputable unofficial-tier dependency under recall-first (medium). The Whop installers, Whop MCP endpoint, and Vercel references are first-party or official sources and are not rated.
MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).
- Direct money access detected (high risk: 0.70). The skill is mainly advisory and diagnostic: choosing a payment network, onboarding, reconciliation, fee comparison, and troubleshooting. It also documents operational routes. It offers business creation through the merchant's user token and connection via Whop MCP/CLI, which it describes as production surfaces. It mentions charges, refunds, and withdrawals as distinct effects and covers recovering or repairing uncertain writes (line 48), including saved-method and recurring-renewal recovery. These are direct financial operations (refunds, charges, withdrawals) that the skill's workflow could carry out through connected tools. The skill does not document an authentication or authorization mechanism enforced by the service layer for those payment routes. It says only that "live operations require supported tools and authority," and it tells the agent to "preserve existing authorization" and not to "silently broaden permissions." Those are agent-followed instructions, not independent access validation. Business creation does use the merchant's user token, which is authenticated, but that covers onboarding and not the charge, refund, or withdrawal routes. The direct operations therefore have no established qualifying control, which puts the skill at high.
Issues (4)
E004
CRITICALPrompt injection detected in skill instructions.
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
W012
MEDIUMUnverifiable external dependency detected (runtime URL that controls agent).
W009
MEDIUMDirect money access capability detected (payment gateways, crypto, banking).
Audit Metadata