The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documents the run_command tool, which permits the agent to execute arbitrary bash commands (e.g., build processes, package installations, and script execution) within the Base44 sandbox environment.
[INDIRECT_PROMPT_INJECTION]: The skill establishes a surface for indirect prompt injection as the agent ingests data from the remote sandbox that could be attacker-controlled.
Ingestion points: File content and directory structures are read from the sandbox via the read_file, grep, and list_directory tools.
Boundary markers: The instructions do not define specific delimiters or instructions to the agent to disregard embedded directives within the files it processes.
Capability inventory: The agent possesses high-privilege capabilities within the sandbox, including writing files (write_file, edit_file) and executing shell commands (run_command).
Sanitization: There is no mention of sanitization or validation for the data retrieved from the sandbox files.
[EXTERNAL_DOWNLOADS]: The skill facilitates fetching configuration and documentation from the vendor's official domain (app.base44.com). It also describes the installation and use of the base44 CLI tool via npx.

base44-remote-dev