The Agent Skills Directory

[PROMPT_INJECTION]: No prompt injection or behavior override patterns detected. The instructions follow natural language patterns for task execution.
[DATA_EXFILTRATION]: No evidence of unauthorized data access or exfiltration. Network activity is limited to official vendor API calls and uses standard placeholder domains like 'example.com' for return URLs.
[OBFUSCATION]: No obfuscated code, encoded strings, or hidden content detected.
[REMOTE_CODE_EXECUTION]: No remote script downloads or dynamic code execution patterns found. The skill relies entirely on predefined vendor tools.
[PRIVILEGE_ESCALATION]: The skill does not attempt to gain elevated system permissions or modify sensitive configuration files.
[PERSISTENCE]: No mechanisms for establishing persistent access (such as cron jobs or shell profile modifications) are present.
[DYNAMIC_EXECUTION]: No runtime code generation or unsafe deserialization techniques are used.
[INDIRECT_PROMPT_INJECTION]: The skill processes data retrieved from the Credyt API, such as product lists and customer details. This constitutes a standard data ingestion surface, but it is used appropriately for verification reporting and does not introduce unusual risks.
[CREDENTIALS_UNSAFE]: No sensitive credentials or API keys are hardcoded. The skill references the public Stripe test card number (4242...) which is standard for simulated payment testing.

billing-verification