The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external sources using web search tools to perform product audits. This creates a surface for indirect prompt injection, where malicious instructions hidden in product reviews or websites could influence the agent's behavior or scoring logic.
Ingestion points: The workflow in Step 1 uses web_search to collect reviews, competitor information, and site data from the public internet.
Boundary markers: The instructions do not define explicit boundary markers or delimiters to separate untrusted search results from the system instructions, nor do they instruct the agent to ignore instructions embedded in the retrieved text.
Capability inventory: The skill utilizes web_search to interact with external web content.
Sanitization: There is no mention of filtering, sanitizing, or validating the content retrieved from web searches before it is used to calculate the 'Purple Cow' remarkability score.

purple-cow-audit