The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted data from multiple external sources, which could contain malicious instructions designed to manipulate the agent's behavior.
Ingestion points: The skill explicitly directs the agent to process external content including customer transcripts, survey results, support tickets, and reviews (in SKILL.md), as well as content from online communities like Reddit, G2, Hacker News, and LinkedIn (in references/source-guides.md).
Boundary markers: The instructions do not define clear delimiters or "ignore embedded instructions" warnings for the ingested data, increasing the risk that the agent might obey instructions found within a customer review or transcript.
Capability inventory: The skill assumes the agent has the capability to read files from the workspace (e.g., .agents/product-marketing.md) and perform web searches to gather intelligence. These capabilities could be misused if triggered by an indirect injection.
Sanitization: There is no evidence of sanitization, filtering, or validation logic provided to handle potentially malicious payloads within the ingested customer feedback or community posts.

customer-research