The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection as it is designed to process and analyze user-provided product documents and external web content.
Ingestion points: User-submitted initiative briefs, PRDs, and roadmaps; web content retrieved via WebFetch, WebSearch, and mcp__scout tools.
Boundary markers: The instructions do not provide explicit delimiters or warnings to the agent to isolate external data from its core logic.
Capability inventory: The skill uses file system reading (Read, Glob, Grep) and network navigation tools. It does not possess file-writing or shell execution capabilities that could be abused for secondary impact.
Sanitization: There is no mention of sanitizing or filtering external content before processing.
[SAFE]: The skill instructions are benign and professional, focusing on product management excellence. No malicious code, prompt injection, or suspicious network activities were identified.

product-expert