The Agent Skills Directory

[SAFE]: The skill demonstrates safe credential management by explicitly prohibiting the agent from requesting API keys, instead directing it to use the official Semrush Model Context Protocol (MCP) for authentication.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data from audited websites via MCP tool outputs (such as issue_details and page_info), creating a standard surface for indirect prompt injection where malicious instructions could be embedded in site content. However, the risk is categorized as safe because the skill is restricted to read-only operations and reporting, and this processing is the primary intended purpose.
Ingestion points: Website audit data retrieved through the issue_details and page_info tools as described in the workflow section of SKILL.md.
Boundary markers: Not explicitly defined in the instructions for separating audit data from agent directives.
Capability inventory: Access is restricted to read-only Semrush MCP tools; no file-writing, unauthorized network exfiltration, or shell execution capabilities are present.
Sanitization: The skill does not define specific sanitization or filtering for the data retrieved from external site audits.

internal-link-audit