The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on npx base44 to perform project management, resource synchronization, and deployment tasks.
[REMOTE_CODE_EXECUTION]: The base44 exec command enables the execution of arbitrary code (read from stdin) against the application's backend environment, which is an intentional feature for administrative scripts and data migrations.
[PRIVILEGE_ESCALATION]: Users with owner or editor roles can use the --privileged flag in base44 exec to bypass Row-Level Security (RLS) rules, providing administrative access to all entity data.
[CREDENTIALS_UNSAFE]: The skill facilitates the management of sensitive application secrets and authentication configurations, including importing credentials from local .env files for social login and SSO providers.
[DYNAMIC_EXECUTION]: The base44 types generate command performs local dynamic code generation by creating TypeScript declaration files based on the project's current resource schemas and updating the project's tsconfig.json file.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to read and process local configuration files, entity schemas, and agent instructions. This creates a surface for indirect prompt injection if these files contain untrusted instructions that influence the agent's behavior.
Ingestion points: Local project files in the base44/ directory (config, entities, agents, functions, skills).
Boundary markers: The instructions do not define explicit delimiters for untrusted content.
Capability inventory: Includes command execution (npx base44), script execution (exec), resource deployment (deploy), and secret management (secrets set).
Sanitization: No explicit sanitization or validation logic is provided for natural language instructions found in project files.

base44-cli