The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to crawl and analyze external websites (e.g., using /content-audit, /technical-seo, and /competitor-gap), which introduces a risk of indirect prompt injection. Malicious instructions hidden in the crawled data could attempt to influence the agent's output or actions.
Ingestion points: External website content retrieved via domain or URL inputs in SKILL.md commands.
Boundary markers: The instructions do not define clear delimiters or warnings to ignore instructions found within the processed external content.
Capability inventory: The skill documentation describes capabilities for file writes (via --export flags) and network crawling operations.
Sanitization: There is no documentation regarding the sanitization or escaping of external web content before it is processed by the agent.
[COMMAND_EXECUTION]: The "Integration Patterns" section provides code examples in JavaScript (execSync) and Python (subprocess.run) that directly interpolate user-controlled variables (like postUrl or domain) into shell commands. This pattern demonstrates a high risk of command injection if implemented literally by the agent or user.
[EXTERNAL_DOWNLOADS]: The installation instructions direct the user to clone the skill from an untrusted GitHub repository (https://github.com/macrohelpclippers/r19-iannuttall-claude-agents-seo.git). While the skill claims to be derived from a known source, the provided repository path is not associated with a recognized trusted organization.

r19-iannuttall-claude-agents-seo