The Agent Skills Directory

[SAFE]: The skill follows security best practices for data analysis agents. It includes explicit instructions to minimize PII, avoid unnecessary data selection (e.g., discouraging SELECT *), and respect access-control boundaries. The logic is contained within local markdown and text-based SQL templates with no executable scripts or remote dependencies.
[INDIRECT_PROMPT_INJECTION]: While the skill ingests external qualitative data (such as support tickets or survey evidence) and behavioral events, it mitigates potential risks through rigorous data quality audits and query validation steps.
Ingestion points: Processes external event history, billing data, and qualitative evidence as described in the SKILL.md workflow.
Boundary markers: The skill relies on strict schema mapping and metric definitions rather than unstructured instruction following for data processing.
Capability inventory: Generating and executing SQL queries against authorized data sources.
Sanitization: Implements a comprehensive "Query validation" protocol in references/guide.md to check join cardinality, time boundaries, and reconciliation to known populations.
[COMMAND_EXECUTION]: The skill provides templates for SQL generation. These are handled as tools within the agent's environment, utilizing parameterization (e.g., :entry_event) to prevent typical SQL injection and ensuring queries use real schema fields and minimal privilege.

user-insights