skills/vm0-ai/vm0-skills/gumroad/Gen Agent Trust Hub

gumroad

Pass

Audited by Gen Agent Trust Hub on Jun 14, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it processes untrusted data from the Gumroad API that could contain malicious instructions designed to influence the agent's behavior.\n
  • Ingestion points: Data retrieved from https://api.gumroad.com/v2/sales, https://api.gumroad.com/v2/products, and https://api.gumroad.com/v2/user (processed via curl in SKILL.md).\n
  • Boundary markers: The instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when the agent processes the retrieved API content.\n
  • Capability inventory: The skill possesses the capability to modify store state, including creating (POST /v2/products), updating (PUT /v2/sales/.../mark_as_shipped), and deleting products (mentioned in the 'When to Use' section).\n
  • Sanitization: No sanitization or validation of the API-returned data is performed before the agent acts upon it.\n- [COMMAND_EXECUTION]: The skill utilizes curl for API interactions and includes a vendor-specific diagnostic command (zero doctor). These operations involve well-known service domains and legitimate vendor resources, consistent with the skill's primary purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 14, 2026, 01:53 AM
Security Audit — agent-trust-hub — gumroad