1password
Warn
Audited by Gen Agent Trust Hub on Jul 7, 2026
Risk Level: MEDIUMCOMMAND_EXECUTIONDATA_EXFILTRATIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Employs
tmuxto manage background sessions and simulate user keystrokes for theop signincommand. This methodology bypasses interactive TTY requirements. - [DATA_EXFILTRATION]: Specifically instructs the agent to access highly sensitive credentials, including SSH private keys and passwords, from 1Password vaults. The output of these commands is captured into the agent's context using
tmux capture-pane. - [EXTERNAL_DOWNLOADS]: Automates the installation of the official 1Password CLI package using the Homebrew package manager.
- [PROMPT_INJECTION]: Creates an attack surface for indirect prompt injection by reading and processing data from external vaults. Malicious content stored within a secret could potentially be interpreted as instructions by the agent.
- Ingestion points:
op read,op inject(SKILL.md, references/cli-examples.md) - Boundary markers: Absent for vault-sourced data.
- Capability inventory: Terminal interaction via
tmux, file writing, and template injection. - Sanitization: Not specified for retrieved secret values.
Audit Metadata