prisma-migration-assistant
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted user inputs, such as database schema definitions and migration names, which are then used as parameters in shell scripts and database queries.
- Ingestion points: User-provided Prisma schema files and migration names in SKILL.md examples and referenced scripts.
- Boundary markers: There are no instructions or markers to distinguish between trusted migration logic and potentially malicious embedded data.
- Capability inventory: The skill uses execSync in TypeScript tests and shell commands in bash scripts to interact with the file system and database.
- Sanitization: The provided script templates do not include validation or sanitization for migration names passed as arguments.
- [DYNAMIC_EXECUTION]: The skill includes TypeScript test templates (tests/migrations/split-name.test.ts) that utilize child_process.execSync to run Prisma CLI commands during the test lifecycle.
- [COMMAND_EXECUTION]: The preview-migration.sh script template takes a positional argument ($1) for the migration name and interpolates it directly into shell commands (npx prisma migrate dev --name "$1" and cat prisma/migrations/*_$1/migration.sql), which represents a potential command injection or path traversal surface if invoked with unsanitized input.
Audit Metadata