The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and act upon data from external, untrusted sources.
Ingestion points: In Phase 1 (Scrape & extract), the skill uses tools such as read_file_content for Google Drive documents and WebFetch or Apify rag-web-browser for live website content to pull raw material for analysis.
Boundary markers: There are no explicit boundary markers or delimiters defined in the instructions to help the agent distinguish between the external data it retrieves and its own internal instructions. It does not include warnings to ignore instructions found within the scraped content.
Capability inventory: The skill has the ability to read files from cloud storage, fetch content from the public web, and write output to Google Sheets in Phase 7.
Sanitization: The skill lacks instructions for sanitizing or filtering the content fetched from external sources before processing it through the generation phases, potentially allowing embedded malicious instructions to influence the agent's behavior.

offer-factory