The Agent Skills Directory

PROMPT INJECTION (SAFE): No instructions to override agent behavior, bypass safety guidelines, or extract system prompts were detected. \n- DATA EXPOSURE & EXFILTRATION (SAFE): No hardcoded credentials, sensitive file paths, or unauthorized network operations were found. The skill correctly advises against leaking internal error details to clients and demonstrates secure handling of authenticated user metadata. \n- OBFUSCATION (SAFE): No Base64, zero-width characters, homoglyphs, or other obfuscation techniques were identified in the text or code snippets. \n- EXTERNAL DOWNLOADS & REMOTE CODE EXECUTION (SAFE): No remote scripts, unauthorized package installations, or dynamic execution patterns were found. The snippets use trusted standard library features and the widely used 'github.com/google/uuid' library. \n- INDIRECT PROMPT INJECTION (SAFE): The skill handles untrusted HTTP data (headers and path variables) using secure, isolated context patterns and provides robust error handling to prevent injection into agent reasoning. \n- SCANNER FALSE POSITIVE (INFO): The automated scanner alert for 'logger.Info' is a false positive. 'logger.Info' is a standard method call in the Go 'slog' package, incorrectly flagged by the scanner due to the '.info' top-level domain.

go-middleware