The Agent Skills Directory

[SAFE]: The skill provides utility classes and implementation examples for defensive security auditing. It follows industry best practices for audit logging, such as using cryptographic hashing to ensure log integrity and tamper-evidence.
[SAFE]: The logging implementation specifically includes privacy-preserving and security-conscious logic, such as masking email addresses and IP addresses, and redacting sensitive keys (e.g., api_key, password, token) from metadata before storage.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data in the form of prompts which creates an indirect prompt injection surface.
Ingestion points: The prompt parameter in KlingAuditClient.text_to_video (SKILL.md) and AuditedKlingAIClient.generate_video (references/audit-wrapped-client.md).
Boundary markers: None identified; prompt data is passed directly to the generation logic and logging functions.
Capability inventory: File system write access for log persistence via the AuditLogger class and network connectivity via the requests library to communicate with api.klingai.com.
Sanitization: The implementation significantly mitigates risk by hashing the prompt content (SHA-256) before writing it to the audit logs, ensuring that the logs themselves do not become a vector for injection persistence.

klingai-audit-logging