The Agent Skills Directory

[PROMPT_INJECTION]: No malicious injection patterns, such as bypass commands, safety filter overrides, or system prompt extraction attempts, were detected. The skill uses role-play in a benign manner to establish expert persona.
[DATA_EXFILTRATION]: The skill does not access sensitive local file paths (e.g., SSH keys, AWS credentials) or environment variables. It suggests checking for project-specific context files like .agents/product-marketing-context.md, which is a standard practice for context-aware agents.
[REMOTE_CODE_EXECUTION]: No remote script downloads, piped execution (curl | bash), or dynamic code evaluation (eval/exec) patterns were found across any of the files.
[COMMAND_EXECUTION]: The skill contains no executable scripts or commands. It is purely instructional markdown and JSON evaluation data.
[EXTERNAL_DOWNLOADS]: References to external services like Semrush, Ahrefs, and specialized AI visibility tools (Otterly, ZipTie) are documented neutrally for monitoring purposes. No automated downloads or package installations are initiated.
[OBFUSCATION]: No obfuscated content, such as Base64 encoded strings, hex escapes, or hidden Unicode characters, was detected in the skill text or metadata.

ai-seo