1password-cli

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests data from 1Password vaults which is considered untrusted external input.
  • Ingestion points: SKILL.md (metadata lookup, secret consumption) and references/cli-usage.md (vault and item discovery).
  • Boundary markers: The skill explicitly instructs the agent to minimize secret exposure, avoid using --reveal flags without consent, and prefer references over plaintext values to keep secrets out of the context where possible.
  • Capability inventory: The skill uses op run for environment injection and op inject for configuration templating, both of which interact with local system processes and the file system.
  • Sanitization: The instructions recommend using --format json and jq to parse metadata safely, ensuring structured data handling.
  • [COMMAND_EXECUTION]: The skill enables execution of local op commands and subprocesses via op run. It provides a detailed shell script template for securely handling service account tokens, including the use of restrictive umask 077, mktemp for secure file creation, and trap for cleanup of sensitive materials.
  • [DYNAMIC_EXECUTION]: The skill utilize dynamic patterns such as eval "$(op signin)" to manage session authentication and generates temporary JSON templates for item creation and editing to avoid passing secrets as command-line arguments.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 12:17 PM
Security Audit — agent-trust-hub — 1password-cli