The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content via the WebFetch tool to detect and validate schema markup.
Ingestion points: The instructions in SKILL.md specify scanning web page source code for JSON-LD, Microdata, and RDFa using the WebFetch tool.
Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores natural language instructions that might be embedded within the structured data being analyzed.
Capability inventory: The skill uses WebFetch and Read tools. Its primary capabilities involve file reading and generating SCHEMA-REPORT.md and generated-schema.json files. It does not exhibit dangerous capabilities like arbitrary command execution.
Sanitization: The skill performs structural and data type validation (e.g., checking for required properties, date formats), but it lacks specific sanitization to filter out prompt injection attacks hidden within the ingested data.

30x-seo-schema