The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill documentation instructs the user/agent to use npx and npm to fetch and run scaffolding tools like create-next-app and create-turbo. Since these tools are maintained by Vercel (a trusted organization), the severity is downgraded to LOW per the [TRUST-SCOPE-RULE].\n- [COMMAND_EXECUTION] (LOW): The SKILL.md file references Python utility scripts (nextjs-init.py and turborepo-migrate.py) for project initialization and migration tasks. These scripts were missing from the uploaded file set, preventing a thorough security audit of their internal command-construction logic and subprocess handling.\n- [REMOTE_CODE_EXECUTION] (LOW): The pattern of executing tools via npx involves fetching and running code from a remote registry at runtime. While the packages mentioned are standard and reputable, the mechanism remains a vector for remote code execution.\n- [PROMPT_INJECTION] (LOW): The skill presents an indirect prompt injection surface as it accepts user-provided parameters (such as project names or paths) that are passed to CLI scripts. \n
Ingestion points: CLI arguments in SKILL.md used by the initialization scripts.\n
Boundary markers: Absent in provided documentation.\n
Capability inventory: Project scaffolding, file system writes, and potential shell command execution.\n
Sanitization: Unverifiable due to the missing source code for the referenced Python scripts.

web-frameworks