The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes scraped business data from external sources through an LLM for classification, creating a surface for indirect prompt injection.
Ingestion points: scripts/classify_leads_llm.py reads scraped lead data (company names, descriptions, and keywords) from JSON files generated by Apify.
Boundary markers: The classification prompt in scripts/classify_leads_llm.py interpolates the scraped data directly into a template without using boundary markers or instructions to ignore potential commands embedded in the descriptions.
Capability inventory: The classification results are used to filter leads and are eventually written to local storage and a Google Sheet via scripts/update_sheet.py.
Sanitization: While company descriptions are truncated to 500 characters, there is no logic to sanitize or filter for prompt injection patterns in the scraped content before it is processed by the LLM.

scrape-leads