The Agent Skills Directory

[SAFE] (SAFE): No malicious behavior patterns detected. The code snippets provide standard educational examples for logging, error tracking, and dependency injection.
[DATA_EXFILTRATION] (SAFE): The skill demonstrates best practices by using environment variables (process.env) for sensitive data such as ENCRYPTION_KEY and SENTRY_DSN rather than hardcoding credentials.
[EXTERNAL_DOWNLOADS] (SAFE): The referenced Node.js packages (inngest, openai, @prisma/client) are standard, well-known libraries for the TypeScript ecosystem.
[COMMAND_EXECUTION] (SAFE): No shell command execution, subprocess spawning, or dangerous system calls were identified.
[PROMPT_INJECTION] (SAFE): No prompt injection or behavior override attempts were found.
[Indirect Prompt Injection] (SAFE):
Ingestion points: Event payloads and function results are processed in transformInput and transformOutput methods.
Boundary markers: None present in the simplified examples.
Capability inventory: Code examples perform logging, metrics tracking, and database updates.
Sanitization: Not explicitly shown, but the context is focused on infrastructure-level middleware rather than LLM prompt construction.
[False Positive Note] (INFO): The automated security alert regarding 'logger.info' being a malicious URL is a false positive; it is a standard logging method call used throughout the software industry.

inngest-middleware