The Agent Skills Directory

[PROMPT_INJECTION]: No evidence of prompt injection attempts, safety filter bypasses, or instruction override patterns were detected. The skill uses instructional language consistent with an expert persona.
[DATA_EXPOSURE]: No hardcoded credentials, sensitive file paths (e.g., SSH keys, AWS configs), or environment variable exfiltration patterns were found. It references standard project context files like .agents/product-marketing.md for background info.
[COMMAND_EXECUTION]: The skill does not contain any executable scripts, shell commands, or instructions for the agent to run potentially harmful system commands.
[REMOTE_CODE_EXECUTION]: There are no patterns involving the download and execution of remote scripts (e.g., curl | bash). It mentions web_fetch and curl solely in the context of their limitations for detecting client-side rendered schema markup.
[OBFUSCATION]: No obfuscation techniques such as Base64 encoding of commands, zero-width characters, homoglyphs, or hidden URLs were detected in the files.
[EXTERNAL_DOWNLOADS]: The skill references well-known and reputable SEO tools and resources (Google Search Console, PageSpeed Insights, Semrush, Ahrefs, etc.) for auditing purposes. No suspicious or untrusted third-party dependencies are requested.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external website data (via web_fetch). While this presents a standard attack surface for indirect prompt injection (where a website being audited could contain instructions targeting the AI), the skill does not grant the agent capabilities that would lead to high-impact exploitation (like writing to local files or exfiltrating credentials based on that data).

seo-audit