The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a workflow for auditing existing schema markup on websites, which introduces a potential surface for indirect prompt injection attacks.\n
Ingestion points: The scripts/schema_validator.py script extracts data from HTML files provided by the user or fetched from the web (e.g., via python3 scripts/schema_validator.py page.html). These files are untrusted external sources.\n
Boundary markers: The skill does not provide the agent with specific delimiters or instructions to ignore potential commands embedded within the extracted schema data, such as instructions hidden in the 'name' or 'description' fields of a JSON-LD block.\n
Capability inventory: The agent has the capability to execute the bundled Python script, read local files, and generate code artifacts based on the results of the audit process.\n
Sanitization: The validation script uses the standard json.loads library for safe parsing of structured data; however, it does not sanitize or filter the string values within the schema objects, which could allow malicious instructions to enter the agent's context.

schema-markup