The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from database tables such as events and system.accounts. This data, which may include user-defined properties or event payloads, represents an injection surface where malicious content could attempt to influence the agent's logic during the analysis or reporting phases.
Ingestion points: Data retrieved from system.accounts, events, and system.account_relationships using the execute-sql tool.
Boundary markers: No specific boundary markers or instructions are provided to the agent to treat queried database values as untrusted data or to ignore embedded instructions.
Capability inventory: The agent has the ability to execute SQL queries, persist findings to a scratchpad via scout-scratchpad-remember, and output results to a report channel using emit_report and edit_report.
Sanitization: The instructions do not include specific sanitization or filtering steps for data retrieved from the database before it is used in report generation or decision-making logic.

signals-scout-customer-analytics