The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill incorporates external research from web searches, video transcripts, and local project context files to inform its advisor simulations. This creates a potential surface for indirect prompt injection where instructions embedded in external content could attempt to influence the agent's behavior.
Ingestion points: Local context files (.agents/product-marketing.md), web search results, and video analysis outputs via external skills.
Boundary markers: The skill defines a rigid output format and 'Grounding Rules' (e.g., no fabricated quotes) to maintain persona integrity, although it lacks specific delimiters to ignore instructions within ingested data.
Capability inventory: The skill reads local files and invokes research tools but does not execute arbitrary shell commands or code from these sources.
Sanitization: The skill relies on grounding advice in provided dossiers as a verification step; however, no explicit sanitization of research content is defined.
[COMMAND_EXECUTION]: The skill provides a feature to create and persist 'Custom Advisor' dossiers by writing markdown files to the .agents/advisors/ directory. This is an intended configuration mechanism and does not involve the creation or execution of executable code or scripts.

marketing-council