The Agent Skills Directory

[REMOTE_CODE_EXECUTION] (HIGH): The skill executes unverified code from the public npm registry using npx get-db. This allows for arbitrary code execution in the user's environment. The risk is significantly increased by the skill's ability to process user-provided SQL files via the --seed flag at runtime.\n- [DATA_EXFILTRATION] (MEDIUM): The skill explicitly reads the sensitive .env file using cat and grep. While it targets the DATABASE_URL key, this mechanism can be exploited to access and expose other secrets stored in the environment file if the agent's context is manipulated.\n- [PROMPT_INJECTION] (HIGH): The skill is highly vulnerable to Indirect Prompt Injection (Category 8) because it ingests untrusted data while maintaining system-level capabilities. \n
Ingestion points: Processes .env files and schema.sql files.\n
Boundary markers: None present; there are no delimiters or instructions to ignore embedded commands within the processed files.\n
Capability inventory: Performs command execution via bash, remote package execution via npx, and package installation via npm install.\n
Sanitization: No validation or sanitization of file content is performed before processing.\n- [EXTERNAL_DOWNLOADS] (MEDIUM): Relies on npx to fetch and execute packages from the npm registry. Neither the package get-db nor the provider 'Neon' are in the strictly defined list of trusted sources.\n- [COMMAND_EXECUTION] (MEDIUM): Extensively uses bash to interact with the local filesystem, environment variables, and external package managers.

neon-instagres