tw-einvoice-guide

Warn

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The reference file references/einvoice-api.md contains hardcoded authentication keys (HashKey="ejCk326UnaZWKisg" and HashIV="q9jcZX8Ib9LM8wYk") and a MerchantID. Although these are documented as credentials for the ECPay staging environment, hardcoding such tokens within skill documentation is an unsafe practice.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input to generate business implementation plans, which creates a potential surface for indirect injection.
  • Ingestion points: The Output Format section in SKILL.md takes dynamic values for business type, transaction volume, and rationale.
  • Boundary markers: No clear delimiters or instructions are provided to the agent to treat user-provided business data as untrusted content.
  • Capability inventory: The skill provides Python code snippets in references/einvoice-api.md that utilize the requests library to perform POST operations to external API endpoints.
  • Sanitization: There is no evidence of sanitization or validation of the user-provided strings before they are interpolated into the final output report.
  • [EXTERNAL_DOWNLOADS]: The skill documents and provides code for interacting with external endpoints at ecpay.com.tw and einvoice.nat.gov.tw. These are well-known and legitimate services for electronic invoicing in Taiwan.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Sep 18, 2026, 02:06 PM
Security Audit — agent-trust-hub — tw-einvoice-guide