The Agent Skills Directory

[COMMAND_EXECUTION]: The initialization script scripts/better_auth_init.py generates TypeScript configuration code by directly interpolating user input from input() calls into code templates. This allows for arbitrary code injection in the resulting auth.ts file if the input is not carefully sanitized.
Evidence: In the _prompt_direct_db method, the db_path variable is placed directly into a string literal within the configuration: f'database: new Database("{db_path}")'. Similarly, the _prompt_mongodb method interpolates db_name directly into the databaseName property.
[DATA_EXFILTRATION]: The better_auth_init.py script attempts to load environment variables from sensitive locations, including .claude/.env and .claude/skills/.env. Reading environment files from these paths can expose the agent's internal configuration and credentials.
Evidence: The _load_env_files method in scripts/better_auth_init.py defines a list of paths including self.project_root / ".claude" / ".env" and iterates through them to parse their contents.
[COMMAND_EXECUTION]: The skill includes an unnecessary binary file scripts/.coverage which contains metadata and absolute file paths from the author's local development environment. Such binary files can be used for obfuscation or to influence the behavior of testing tools like coverage.py.
Evidence: Hex analysis of the .coverage file reveals a SQLite database containing local paths such as /mnt/d/www/claudekit/claudekit-engineer/.claude/skills/better-auth/scripts/better_auth_init.py.

better-auth