frappe-bench-cli

Warn

Audited by Gen Agent Trust Hub on Jun 14, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCREDENTIALS_UNSAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill enables high-privilege system and application management through the bench CLI.
  • bench console provides an interactive Python shell with the Frappe environment loaded.
  • bench mariadb provides direct access to the database shell.
  • bench execute allows the execution of arbitrary Python functions within the context of an app.
  • [EXTERNAL_DOWNLOADS]: The bench get-app command is used to download source code from remote repositories. While the example uses GitHub, the command supports arbitrary URLs.
  • [REMOTE_CODE_EXECUTION]: The workflow of downloading external code (get-app) followed by installation (install-app) or execution (migrate, execute) allows for the execution of unverified remote code.
  • [CREDENTIALS_UNSAFE]: Several administrative tasks involve sensitive data handling.
  • bench set-admin-password <pwd> accepts passwords as plain-text command-line arguments, which typically remain in the user's shell history.
  • bench set-config is used to write secrets directly to configuration files.
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection (Category 8).
  • Ingestion points: Site hostnames, application names, and Git URLs provided to commands like new-site or get-app (SKILL.md).
  • Boundary markers: None identified; untrusted inputs are interpolated directly into shell commands.
  • Capability inventory: Subprocess execution via bench, database access via mariadb, and arbitrary Python execution via console/execute (SKILL.md).
  • Sanitization: No evidence of input validation or escaping for the parameters passed to the bench CLI.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Jun 14, 2026, 11:50 PM