data-validation
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill is authored with a strong security posture, explicitly forbidding any data modification (INSERT/UPDATE/DELETE), schema changes, or connections to production environments. It requires manual human review or explicit user requests for report generation.\n- [DATA_EXFILTRATION]: The skill resolve data sources by reading project-local configuration files such as
.env.testor.env.localto identifyDATABASE_URL. This activity is restricted to non-production targets declared within the repository and does not involve external transmission of the discovered credentials.\n- [COMMAND_EXECUTION]: Permission is granted to execute the project's existing read-only data scripts, provided they are invoked with an explicit dry-run flag to prevent side effects.\n- [PROMPT_INJECTION]: The skill is subject to potential indirect prompt injection as it processes and samples data from external sources like database tables and fixture files. The skill mitigates this surface through structured reporting and a mandatory data sanitization phase.\n - Ingestion points: Project-local fixtures (
testdata/,fixtures/,seeds/) and database records accessed viaSELECTorEXPLAINqueries.\n - Boundary markers: Findings are strictly encapsulated within a predefined markdown report template with specific markers for attribution.\n
- Capability inventory: Local file system read access and execution of shell scripts within the repository context.\n
- Sanitization: A mandatory PII masking protocol is enforced for all sampled data, covering emails, phone numbers, names, addresses, and sensitive identification numbers.
Audit Metadata