make-video

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 Atmee clip maker that asks before paid calls, and its upload and install paths stay on the vendor's own domains and repo, but it persistently widens the user's permissions (auto-allowing atmee tools in settings.json and relaxing the Codex sandbox network setting), which goes beyond what the task strictly needs.

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

  • Direct money access detected (medium risk: 0.30). Operation: generate_clip (line 135) is a paid call that debits the creator's prepaid credit balance per clip, and the skill asks for an explicit yes and then sends confirm: true (lines 62-70, 223-225). This is closer to instructing a debit of a funded balance than to a pure service-cost exclusion, so I treat it as in scope under the recall-first override. It is a direct operation because it moves value out of the creator's account. Control: the service enforces sign-in and API-key access for the account (get_api_key, get_account, and get_balance in the allowed tools, lines 7 and 129-130; sign-in check at lines 146-148; accountId routing at lines 123-125). That service-side authentication gates the paid route. The agent-followed price confirmation and the confirm: true flag are not counted as the control. This is a borderline classification, since a reader could instead treat it as operating a paid API.

Issues (2)

E004
CRITICAL

Prompt injection detected in skill instructions.

W009
MEDIUM

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

Audit Metadata
Risk Level
CRITICAL
Analyzed
Nov 19, 2026, 10:58 PM
Issues
2