The Agent Skills Directory

[SAFE]: The skill provides informational content and documentation for the Unbounce platform.
[DATA_EXPOSURE]: The skill reads and writes to references/learnings.md to maintain persistent context between interactions. This data is stored locally within the skill's reference directory and does not involve external transmission of sensitive information.
[PROMPT_INJECTION]: The skill uses references/learnings.md as an ingestion point for accumulated knowledge which represents a surface for indirect prompt injection. • Ingestion points: references/learnings.md (read at start of session). • Boundary markers: Absent. • Capability inventory: The skill can append content to the local learnings file. • Sanitization: No sanitization is performed on user-provided tips before storage.

sales-unbounce