The Agent Skills Directory

[METADATA_POISONING]: The file scripts/.coverage is a binary SQLite database containing code coverage data from the author's development environment. It includes absolute file paths (e.g., /mnt/d/www/claudekit/...), which exposes the author's local directory structure but does not affect the safety of the skill for the end user.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface through user-supplied project names and existing monorepo metadata.
Ingestion points: The name argument in nextjs_init.py and package definitions in turborepo_migrate.py.
Boundary markers: The validate_name method ensures the project name is alphanumeric.
Capability inventory: Scripts perform directory creation and file writing using standard libraries.
Sanitization: Explicit validation in nextjs_init.py prevents path traversal via project names.
[COMMAND_EXECUTION]: The skill documentation suggests standard industry practices like npx create-next-app, which are safe when run in a local development environment. The included Python scripts use standard libraries to manage file systems and do not contain hidden execution vectors.

web-frameworks