The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's core functionality involves crawling up to 500 pages of external, untrusted websites using scripts/fetch_page.py. This content is then processed by multiple subagents (e.g., seo-content, seo-technical, seo-geo) to generate a health score and reports. There is a lack of explicit boundary markers or instructions to ignore embedded malicious prompts within the crawled HTML/metadata, which could theoretically influence the agent's analysis or report output.
Ingestion points: Site crawling functionality and homepage fetching in scripts/fetch_page.py and the main audit process.
Boundary markers: No specific delimiters or "ignore" instructions are defined for the crawled content.
Capability inventory: The skill executes local Python scripts, delegates to subagents, and writes files (FULL-AUDIT-REPORT.md, ACTION-PLAN.md, and PDF reports).
Sanitization: No sanitization or filtering of the crawled external content is mentioned.
[COMMAND_EXECUTION]: The skill workflow involves executing several local Python scripts to handle specialized tasks, including authentication checks and report generation.
Evidence: Instructions include running python scripts/google_auth.py --check, python scripts/backlinks_auth.py --check, python scripts/drift_history.py <url>, and python scripts/google_report.py --type full.

seo-audit