The Agent Skills Directory

[SAFE]: A thorough review of all skill components, including the 7 specialist agent instructions, reveals no evidence of malicious patterns, prompt injection, or obfuscation. The skill adheres to best practices for AI agent instruction design.
[INDIRECT_PROMPT_INJECTION]: The system is designed to ingest a brand-context.md file from the user's project to align its output with company branding. This creates a potential surface for indirect prompt injection; however, the skill explicitly instructs the agent to treat this as binding instructional context with specific rules, and it does not grant the content of that file any executable or privileged capabilities.
[EXTERNAL_DOWNLOADS]: Several external URLs are referenced throughout the documentation to provide the agent or user with technical background (e.g., Google Patents, Cloudflare's Content Signals Policy, and various SEO playbooks hosted on GitHub by the author, 'shalintripathi'). None of these references involve remote code execution or automated script downloads.
[COMMAND_EXECUTION]: The skill does not contain any shell commands, dynamic context injections, or scripts that would interact with the underlying operating system in a privileged or dangerous manner.
[METADATA_POISONING]: Although the skill's description field in the YAML frontmatter is dense with keywords, they are relevant to the SEO domain and do not contain hidden instructions or malicious patterns intended to poison the agent's lifecycle.

seo-growth