The Agent Skills Directory

[SAFE]: The skill demonstrates high security awareness by implementing comprehensive host and IP validation in its Python scripts (analyze_landing.py, capture_screenshot.py, and fetch_page.py). These scripts explicitly check for private, loopback, and reserved IP ranges to prevent SSRF attacks during landing page analysis.
[SAFE]: The multi-agent orchestration logic used for parallel auditing is standard and does not exhibit any privilege escalation or persistence mechanisms.
[SAFE]: No hardcoded credentials or sensitive file access patterns were identified. The tool correctly uses placeholders and standard API references for advertising platforms.
[SAFE]: Although the skill processes external web content, which is a potential surface for indirect prompt injection, it is inherent to the skill's primary function (auditing websites). The potential risk is mitigated by the agent's internal safety layers.
Ingestion points: The scripts analyze_landing.py and fetch_page.py ingest untrusted data from external URLs, including page titles, H1 headers, meta descriptions, and schema markup.
Boundary markers: Not explicitly defined within the scripts' data structures, relying on the calling agent's context management.
Capability inventory: The skill can fetch web content, capture screenshots, and coordinate specialized sub-agents for advertising platform audits.
Sanitization: The skill performs rigorous network-level sanitization (SSRF guards) before ingestion, though it does not modify the retrieved text content itself.

ads