The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection due to its processing of external research data and user-provided persona definitions.- Ingestion points: External data enters the agent context via the 'Live Research Pass' (web search or external research tools) and through user-supplied data for 'Custom Advisor' dossiers (SKILL.md).- Boundary markers: Absent. The instructions do not define specific delimiters or 'ignore embedded instructions' warnings to prevent the agent from obeying malicious commands hidden in research results.- Capability inventory: The skill allows writing markdown files to the '.agents/advisors/' directory for persistence (SKILL.md). No high-risk capabilities such as arbitrary command execution, network exfiltration, or privilege escalation were detected.- Sanitization: Absent. There is no logic provided for sanitizing, validating, or filtering the content retrieved during the research pass or provided by users for custom dossiers.

suede-marketing-council