clerk-migration-deep-dive
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFEDYNAMIC_CONTEXT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_CONTEXT_INJECTION]: The skill utilizes a dynamic context command
!npm list...in theSKILL.mdfile. This command executes at load time to detect the presence of specific authentication packages in the local environment and display them to the agent. - [COMMAND_EXECUTION]: The instructions include shell commands for exporting data via
curlfrom Auth0's Management API and managing dependencies usingnpm. These operations are standard for the stated migration task. - [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for ingesting untrusted external data by reading user profiles from JSON files exported from providers like Auth0 or Firebase.
- Ingestion points: Data is read from
auth0-users.jsonandnextauth-users.jsoninscripts/import-to-clerk.tsandscripts/update-db-references.ts. - Boundary markers: None identified in the provided scripts.
- Capability inventory: The skill has the capability to write to the local file system, update a database via Prisma, and perform network requests to the Clerk API.
- Sanitization: The scripts perform field mapping but lack explicit sanitization or validation of the ingested user data before processing.
Audit Metadata