skills/breakthrough-edu/breakthrough-database-skills/breakthrough-database-to-dashboard/Gen Agent Trust Hub
breakthrough-database-to-dashboard
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from external sources including Airtable, Notion, Google Sheets, and local CSV/Excel files as described in
SKILL.mdandreferences/getting-the-data.md. - Ingestion points: Data is introduced into the agent context via URL links or direct file uploads.
- Boundary markers: The instructions lack explicit directives for the agent to treat data records as untrusted or to ignore instructions embedded within the database content (e.g., in column metadata or cell values).
- Capability inventory: The agent has the capability to execute code for data processing (Iron Rule R2) and write files/artifacts (Phase 4).
- Sanitization: While Iron Rule R7 mandates the removal of personal identifiers (PII), there are no instructions to escape or sanitize data strings to prevent subversion of the agent's logic or to prevent Cross-Site Scripting (XSS) in the generated dashboard.
- [DYNAMIC_EXECUTION]: As defined in
references/building-the-dashboard.md, the skill generates a standalone.htmlartifact by embedding data slices and precomputed aggregates into a template. This is a standard functional capability but involves the dynamic creation of executable web content from user-provided data inputs.
Audit Metadata