The Agent Skills Directory

[Unverifiable Dependencies & Remote Code Execution] (MEDIUM): The skill references and executes several shell scripts (e.g., scripts/init.sh, scripts/db-migrate.sh, scripts/db-reset.sh) whose contents are not provided for analysis. Executing opaque local scripts can lead to arbitrary code execution if those files are modified.
[Indirect Prompt Injection] (HIGH): This skill demonstrates a high-risk capability tier by ingesting untrusted data (external templates and scripts) and possessing write/execute capabilities.
Ingestion points: Template files located in templates/ and script files in scripts/ are copied and executed.
Boundary markers: Absent. The agent is instructed to copy and run these files without verification.
Capability inventory: Extensive subprocess execution including bun create, bun add, docker compose, and bash script execution.
Sanitization: Absent. There is no validation or escaping of the content within the template files before they are written to the filesystem or executed.
[Command Execution] (LOW): The skill makes extensive use of system commands like docker compose, mkdir, cp, and openssl. While appropriate for a setup tool, these commands operate with the privileges of the executing user.

nextjs-fullstack-setup