The Agent Skills Directory

[SAFE]: The skill consists entirely of instructional Markdown content and evaluation data. It does not contain executable scripts, subprocess calls, or system commands.
[PROMPT_INJECTION]: The skill instructs the agent to read local context files (e.g., .agents/product-marketing.md). This creates a surface for indirect prompt injection from those local files. However, because the skill does not possess high-privilege capabilities (such as code execution, file system writes, or network exfiltration), the risk of exploitation is negligible. Evidence chain for this surface: (1) Ingestion points: SKILL.md reads .agents/product-marketing.md; (2) Boundary markers: Absent; (3) Capability inventory: Mentions using tools like semrush and ahrefs for SEO data (read-only); (4) Sanitization: Absent.
[EXTERNAL_DOWNLOADS]: All external references point to reputable documentation (e.g., Google Developers, llmstxt.org) or well-known SEO industry tools. No patterns of remote script execution or unauthorized downloads were found.
[DATA_EXFILTRATION]: No hardcoded credentials, sensitive file path access, or network calls to untrusted domains were identified.

ai-seo