The Agent Skills Directory

[COMMAND_EXECUTION]: Provides instructions for executing powerful administrative commands such as bench drop-site --force for deleting data and sudo bench setup production for system-level configuration.
[COMMAND_EXECUTION]: Documents the use of bench execute and bench console, which allow for the execution of arbitrary Python code within the application environment.
[COMMAND_EXECUTION]: Identified an indirect prompt injection surface where the skill allows ingesting untrusted data (via bench data-import and bench import-csv in references/commands.md) and provides high-capability tools like bench execute and bench console without documenting specific sanitization or boundary markers.
[REMOTE_CODE_EXECUTION]: Details the installation of applications from remote repositories using the bench get-app command, which fetches and prepares code for execution.
[CREDENTIALS_UNSAFE]: Identifies the location and schema of site_config.json and common_site_config.json, which contain sensitive database credentials and configuration secrets.
[DATA_EXFILTRATION]: Documents the bench ngrok command, which allows exposing local site instances to the public internet, potentially bypassing network security controls.
[EXTERNAL_DOWNLOADS]: References the acquisition of software updates and applications from well-known services such as GitHub and official package registries.

frappe-ops-bench