The Agent Skills Directory

[SAFE]: No malicious code or suspicious behaviors were detected. The skill scripts (scripts/validate.py and scripts/visible_content_match.py) use standard Python libraries and safe regex-based parsing to validate structured data without requiring external dependencies or network access.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data (web page content) to generate structured markup.
Ingestion points: The page_url_or_page_content field in contract.yaml is the primary entry point for untrusted external data.
Boundary markers: SKILL.md contains explicit instructions to ignore any external content that is not visible to the user and to refuse fabrication, serving as a behavioral boundary against injection attempts.
Capability inventory: The skill's capabilities are restricted to reading files and executing local validation scripts. It lacks the ability to write to sensitive system paths or perform network exfiltration.
Sanitization: The visible_content_match.py script actively sanitizes input by removing <script>, <style>, and <noscript> tags before comparing schema values against page text.

seo-schema-builder