The Agent Skills Directory

Indirect Prompt Injection Surface: The skill ingests external research data via a JSON payload. This is a standard design for data-processing skills and includes several safety measures to manage the ingestion surface.
Ingestion points: Data enters the skill through the governed.json file as defined in SKILL.md.
Boundary markers: The instructions require validation against the rtl.consumer-insight-synthesis.v1 contract schema before any synthesis occurs.
Capability inventory: The skill is focused on document generation (INSIGHT-<topic>.md) and does not have access to network tools, subprocess execution, or sensitive file system paths.
Sanitization: Inputs are governed by the strict schema validation defined in contracts/rtl.consumer-insight-synthesis.v1.json.
Data Privacy and Governance Logic: The skill incorporates business logic to verify study licenses, geographical restrictions, and expiration dates. These checks are designed to prevent the unauthorized inclusion of restricted data in the final artifacts, which is a positive security and compliance practice.
File System Operations: The skill's output is limited to creating markdown reports and provenance records based on the processed data. There are no patterns indicating attempts to access credentials, system configurations, or other sensitive files.

insight-brief