The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves data from external MongoDB collections, which serves as a source of untrusted content that could contain embedded malicious instructions.
Ingestion points: Data is ingested through multiple database queries defined in SKILL.md, including db.users.findOne(), db.bases.find(), and various aggregation pipelines.
Boundary markers: The instructions lack delimiters or specific warnings for the agent to distinguish between data and instructions when processing database records.
Capability inventory: The skill utilizes the Bash tool to execute mongosh commands, providing a pathway for executing logic based on retrieved data.
Sanitization: There is no evidence of filtering, escaping, or validation of the database content before it is processed by the agent.

db-explore