The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from user prompts and project files, which is then processed by simulated marketing advisors. The surface is well-defined and associated with the primary informational purpose of the skill.
Ingestion points: User-provided marketing questions, project context in .agents/product-marketing.md, and external content from live research tools.
Boundary markers: The instructions include a non-negotiable 'simulation' label for all outputs to provide context, though specific technical delimiters for input data are not present.
Capability inventory: The skill uses file-read capabilities and optional research tools to inform persona simulations; it lacks dangerous execution or exfiltration tools.
Sanitization: No specific input sanitization is defined, as the skill relies on persona grounding to filter and frame the response.
[SAFE]: The custom advisor feature manages project-level advisor dossiers in the .agents/ folder, which is a standard method for configuration management and does not involve suspicious persistence or privilege escalation tactics.

marketing-council