The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection risk surface detected. The skill aggregates data from a knowledge base and user-provided descriptions to generate content and update internal entities.
Ingestion points: Data enters the context via get_entity, search_knowledge_base, and get_playbook tools, as well as user-defined descriptions in the /octave:launch command flow.
Boundary markers: The instructions interpolate retrieved data (e.g., <product details, launch positioning>, <full library context>) directly into generation tools like generate_content and generate_email without using protective delimiters or instructions to ignore embedded commands.
Capability inventory: The skill utilizes create_entity, update_entity, and generate_email, providing significant write and outbound communication capabilities that could be influenced by malicious data.
Sanitization: There is no evidence of sanitization or validation of retrieved library content before its use in generation or update functions.

launch